You are on page 1of 30

Journal of Research in Personality 35, 138–167 (2001

)
doi:10.1006/jrpe.2000.2302, available online at http://www.idealibrary.com on
Refining the Architecture of Aggression: A Measurement Model
for the Buss–Perry Aggression Questionnaire
Fred B. Bryant and Bruce D. Smith
Loyola University Chicago
Among the most popular measures of aggression is the 29-item, self-report Ag-
gression Questionnaire (AQ; Buss & Perry, 1992; Buss & Warren, 2000). Structural
analyses of the AQ have revealed four underlying factors: Physical Aggression,
Verbal Aggression, Anger, and Hostility. However, these four factors explain too
little common variance (i.e., about 80%) to be an adequate measurement model. In
the present study, we used confirmatory factor analysis with a total sample of 1154
respondents to compare four alternative measurement models for the AQ that are
currently in use. Replicating earlier work, none of these models fit the data well,
and the original four-factor model achieved only mediocre goodness-of-fit in three
independent samples (GFI ϭ .76 Ϫ .81). To develop a more appropriate measure-
ment model, we omitted items with low loadings or multiple loadings based on
principal components analysis and excluded items with reverse-scored wording.
This yielded a 12-item, four-factor measurement model with acceptable goodness-
of-fit (GFI ϭ .94). Secondary analysis of two independent data sets confirmed the
refined model’s generalizability for British (Archer, Holloway, & McLoughlin,
1995; GFI ϭ .93) and Canadian (Harris, 1995; GFI ϭ .94) samples. The refined
model yielded equivalent factor structures for males and females in all three sam-
ples. We also replicated the refined four-factor model in two additional American
samples, who completed a new short form of the AQ containing only the subset of
12 items in random order. Additional analyses provided evidence supporting the
model’s construct validity and demonstrated stronger discriminant validity for the
refined Hostility factor compared to its predecessor. The new short form of the AQ
thus not only contains fewer than half as many items as the original, but also is
psychometrically superior. © 2001 Academic Press
The authors thank John Archer and Julie Harris for graciously providing us with data from
their published articles. We also gratefully acknowledge the helpful advice of Mary Harris,
the invaluable research assistance of Rebecca Guilbault, and the insightful editorial feedback
of Craig Colder and an anonymous reviewer. Earlier versions of this article were presented
at the American Psychological Association convention, Chicago, IL, August 1997, and at the
Joint Meeting of the Classification Society of North America and the Psychometric Society,
Urbana, IL, June 1998.
Address correspondence and reprint requests to Fred B. Bryant, Department of Psychol-
ogy, Loyola University Chicago, 6525 North Sheridan Road, Chicago, IL 60626. E-mail:
fbryant@luc.edu.
138
0092-6566/01 $35.00
Copyright © 2001 by Academic Press
All rights of reproduction in any form reserved.
ARCHITECTURE OF AGGRESSION 139
Much work has emphasized the role of physical aggression, verbal aggres-
sion, anger, and hostility as subtraits in a global conceptualization of aggres-
sion (Buss, 1961; Buss & Durkee, 1957; Buss & Perry, 1992; Harris, 1995;
Zillmann, 1979). Early measurement of aggression used an experimental
methodology, required a laboratory, and suffered difficulties in interpreting
aggressive intent (Zillmann, 1979). To reduce the time, effort, and resources
involved in measuring aggression, Buss and Durkee (1957) developed a 75-
item self-report instrument, the Buss–Durkee Hostility Inventory, which ag-
gression researchers have often used.
To improve the psychometric properties of this instrument, Buss and
Perry (1992) more recently developed a 29-item self-report questionnaire,
the Aggression Questionnaire (AQ; Buss & Perry, 1992; Buss & Warren,
2000). They designed the AQ to measure four dispositional subtraits of ag-
gression, which they defined as follows: ‘‘Physical and verbal aggression,
which involve hurting or harming others, represent the instrumental or mo-
tor component of behavior. Anger, which involves physiological arousal
and preparation for aggression, represents the emotional or affective com-
ponent of behavior. Hostility, which consists of feelings of ill will and in-
justice, represents the cognitive component of behavior’’ (Buss & Perry,
1992, p. 457).
In constructing this questionnaire, Buss and Perry (1992) borrowed some
items intact from the earlier Hostility Inventory, revised other Buss–Durkee
items to improve clarity, and added many new items to generate an initial
pool of 52 questions. They then administered this set of 52 questions to
three successive samples of 406, 448, and 399 college students and analyzed
the structure of responses using exploratory principal components analysis
with oblique rotations. Although they had originally generated items for six
a priori components of aggression (Physical Aggression, Verbal Aggression,
Anger, Indirect Aggression, Resentment, and Suspicion), only four corre-
lated factors emerged—Physical Aggression, Verbal Aggression, Anger, and
Hostility—on which a core set of 29 items loaded, and this four-factor struc-
ture appeared to replicate across all three samples. Buss and Perry (1992)
next used confirmatory factor analysis to evaluate the goodness-of-fit of three
alternative measurement models for the set of 29 items: (a) a global one-
factor model that assumes all items reflect a single general aggression factor;
(b) a four-factor model that represents the structure found in the principal
components analysis; and (c) a hierarchical factor model that assumes the
four correlated, first-order factors reflect a single, second-order ‘‘super fac-
tor’’ of aggression. Figure 1 presents a graphical representation of each of
these three measurement models.
As a sole measure of each model’s goodness-of-fit, Buss and Perry (1992)
computed the ratio of chi-square to degrees of freedom (cf. Hoelter, 1983).
Based on the notion that ratios under 2 reflect acceptable fit, Buss and Perry
FIG. 1 Three measurement models currently in use for the 29-item Aggression Question-
naire (AQ; Buss & Perry, 1992). Squares represent measured variables (or AQ items), and
circles represent latent constructs (or AQ factors). Arrow-headed straight lines connecting
latent constructs to measured variables represent item factor-loadings (λs). Two-headed,
curved lines connecting latent factors represent factor interrelationships (φs). The small, arrow-
headed straight line to each measured variable represents unique variance associated with
measurement error, or the joint effect of unmeasured influences and random error (θ
δ
). The
unidimensional model (top) assumes that a single, first-order factor (Global Aggression) ex-
plains the covariation among the 29 AQ items. The multidimensional model (center) assumes
that four, interrelated first-order factors (Physical Aggression, Verbal Aggression, Anger, and
Hostility) explain the covariation among the 29 AQ items. The hierarchical model (bottom)
assumes that a single, global second-order factor (Global Aggression) underlies the covariation
among the four first-order factors. The small, arrow-headed straight line to each first-order
latent factor in the hierarchical model represents specific variance that is unrelated to the
second-order latent factor (ψ).
ARCHITECTURE OF AGGRESSION 141
(1992) concluded that both the four-factor model (χ
2
/df ϭ 1.94) and the
higher-order factor model (χ
2
/df ϭ1.95) fit the pooled data reasonably well,
whereas the one-factor model (χ
2
/df ϭ2.27) did not. Recalculating the model
chi-squares from the known degrees of freedom reveals that both the four-
factor model, χ
2
(371) ϭ719.7, and the higher-order model, χ
2
(373) ϭ727.4,
fit the data significantly better (both ∆χ
2
s Ͼ 128.3, ps Ͻ .00001) than did
the one-factor model, χ
2
(377) ϭ 855.8. Because Buss and Perry (1992) re-
ported no other goodness-of-fit measures, however, we know nothing about
the proportion of variance–covariance information that these models ex-
plained in their data or how much better these models fit relative to a worse-
case ‘‘null’’ model that assumes there are no common factors. Such measures
of absolute and relative fit would be useful in deciding whether either of the
multidimensional frameworks represents an acceptable measurement model
for the AQ.
Although Buss and Perry (1992) proposed their four-factor solution as a
measurement model for the AQ, more recent analyses (Archer, Kilpatrick, &
Bramwell, 1995; Harris, 1995; Williams, Boyd, Cascardi, & Poythress,
1996) suggest that this structure explains too little common variance among
the 29 items (i.e., about 80%) to serve as a measurement model. Accordingly,
the primary goal of the present study was to develop an acceptable measure-
ment model for the AQ and to assess its construct validity.
As a means of improving the measurement precision of the AQ, previous
researchers have proposed discarding AQ items that are relatively unreliable
indicators of Hostility. Indeed, omitting Buss and Perry’s (1992, Table 1,
p. 454) sixth and eighth indicators of Hostility has been found to increase
the reliability of the Hostility factor in both Canadian (Harris, 1995) and
Dutch (Meesters, Muris, Bosma, Schouten, & Beuving, 1996) samples. Yet,
this approach is not without its critics (Bernstein & Gesn, 1997). The key
theoretical issue here is whether it is better to have a global, somewhat het-
erogeneous construct of known theoretical utility or to have a more specific
and psychometrically purified construct (cf. Bryant, Yarnold, & Grimm,
1996).
Researchers using the AQ have typically adopted one of two dominant
strategies for scoring the instrument. The first strategy assumes that all AQ
items reflect a single underlying construct reflecting a person’s global predis-
position toward aggression. With this unidimensional approach, researchers
simply sum responses to the 29 items to construct an AQ total score (e.g.,
Buss & Perry, 1992). The second strategy assumes that aggression consists
of four correlated dimensions reflecting a person’s predisposition toward ag-
gression in the physical, verbal, emotional (Anger), and cognitive (Hostility)
domains. With this multidimensional approach, researchers construct four
separate subscale scores by summing or averaging responses to the set of
AQ items tapping each domain of aggression (e.g., Felsten & Hill, 1998).
142 BRYANT AND SMITH
Yet, neither of these two approaches adequately captures the variation in
people’s responses to the AQ. More specifically, measurement error and
other constructs in addition to aggression have an unacceptably large influ-
ence on responses to the set of 29 AQ items. Although the theoretical model
underlying the AQ is conceptually well grounded in the aggression literature,
researchers need a better operational framework for quantifying responses
to the instrument. With this aimin mind, we sought to improve the correspon-
dence between the conceptual and operational definitions underlying the AQ
and to develop a more reasonable measurement model for the AQ.
We worked with five independent data sets: three primary data sets that
we collected for this study and two archival data sets that others had collected
earlier. First, we collected a new data set to evaluate and compare the explan-
atory power of four different measurement models for the AQ that are cur-
rently in use in the literature. We also used these new data to develop a
better-fitting refined version of Buss and Perry’s (1992) four-factor model
and to compare the convergent and discriminant validity of this refined model
with that of the original. We then obtained two preexisting AQ data sets—
a British sample (Archer, Holloway, & McLoughlin, 1995) and a Canadian
sample (Harris, 1995)—with which we assessed the cross-sample generaliz-
ability of both the original and refined models. Finally, we collected new
data from two additional American samples to evaluate the refined model’s
replicability using a new ‘‘short form’’ of the AQ, containing only the subset
of 12 items in random order.
METHOD
Participants and Procedure
Sample 1. The first sample consisted of new data from 307 American undergraduates (173
females, 131 males, and 3 who did not report gender) at a private metropolitan university who
voluntarily participated. Average age was 18.94 (SDϭ1.21). Respondents completed a battery
of tests, including the 29-item Aggression Questionnaire (AQ; Buss & Perry, 1992; Buss &
Warren, 2000).
Sample 2. The second sample, which we used for cross-validation, was originally collected
by Archer, Holloway, and McLoughlin (1995). The sample consisted of 200 British undergrad-
uates (100 females and 100 males). Average age was 25.13 (SD ϭ 6.17). Participants com-
pleted the 29-item AQ using the same 5-point scale. The data consisted of the raw AQ data
analyzed by Archer et al. (1995).
Sample 3. The third sample, also used for cross-validation, was originally collected by Harris
(1995). The sample consisted of 306 Canadian undergraduates (151 female and 155 male).
Average age was comparable to U.S. college samples (cf. Harris, 1995). Participants completed
the 29-item AQ using the same 5-point scale. The data consisted of the covariance matrix
analyzed by Harris (1995).
Sample 4. The fourth sample was used to assess the replicability of the refined four-factor
model using a new ‘‘short form’’ of the AQ. The sample consisted of 171 American undergrad-
uates (123 females and 48 males) at a private metropolitan university who voluntarily partici-
pated. Average age was 18.35 (SD ϭ 0.84). Respondents completed a shortened version of
the AQ containing only the 12 items comprising the refined four-factor measurement model.
ARCHITECTURE OF AGGRESSION 143
Sample 5. The fifth sample was also used to assess the replicability of the refined four-
factor model using the new ‘‘short form’’ of the AQ. The sample consisted of 170 American
undergraduates (124 females and 46 males) at a private metropolitan university who volun-
tarily participated. Average age was 18.64 (SD ϭ1.57). Respondents completed the shortened
12-item version of the AQ.
Aggression Questionnaire
Original AQ. Respondents in Samples 1–3 completed a battery of tests, including the 29-
item Aggression Questionnaire (Buss & Perry, 1992; Buss & Warren, 2000). Participants rated
how well each AQ item described themselves using the original 5-point scale, ranging from
extremely uncharacteristic of me (1) to extremely characteristic of me (5), as defined by the
original instrument. Following Buss and Perry’s (1992, p. 453) instructions, each sample re-
ceived a different random ordering of the 29 AQ items.
Short form of the AQ. Respondents in Samples 4 and 5 completed a new, shortened version
of the AQ containing only the 12 items comprising the refined four-factor measurement model.
Using Buss and Perry’s (1992, p. 454) Table 1, the randomized order of the items was 11,
23, 8, 25, 21, 14, 15, 2, 13, 24, 6, and 20. Participants rated each AQ item using a 6-point
scale, ranging from extremely uncharacteristic of me (1) to extremely characteristic of me
(6). Changing from the original 5-point scale to a 6-point scale eliminated the scale’s midpoint,
thereby forcing respondents to decide whether each statement was characteristic of them. A
response scale with an even number of points also better enables researchers to use a median-
split on single items to categorize respondents as aggressive versus nonaggressive in specific
situations (cf. Sudman & Bradburn, 1982). As a precedent, Velicer, Govia, Cherico, and Corri-
veau (1985) have modified the response scale of the Buss–Durkee Hostility Scale to make
this instrument more reliable.
Criterion Measures
In addition to completing the AQ, a random subset of 180 participants in Sample 1 (70
males and 110 females) also filled out a set of criterion measures for use in evaluating the
AQ’s construct validity. These criterion measures served as standards for assessing the conver-
gent and discriminant validity of the dimensions comprising both Buss and Perry’s original
four-factor model and the new, refined measurement model for the AQ.
Physical Aggression. As a criterion measure of physical aggression, we used the Assault
subscale from the Buss–Durkee Hostility Inventory (Buss & Durkee, 1957). Buss (1961) re-
ported a 5-week test–retest reliability of .78 for this subscale. As validity evidence, men who
have committed domestic violence score higher on the Assault subscale compared to controls
(Maiuro, Cahn, Vitaliano, Wagner, & Zegree, 1988). Although the original Buss–Durkee As-
sault subscale consisted of 10 items, we decided to use only those Assault items that had not
been adapted by Buss and Perry in constructing the AQ. Specifically, 5 Assault subscale items
were worded almost identically in Buss and Perry’s AQ (Buss–Durkee items 9, 17, 25, 65,
and 70). We chose to exclude these items from the Assault subscale because their comparable
wording might otherwise spuriously inflate the degree of association between this criterion
measure and the AQ (cf. Nichols, Licht, & Pearl, 1982).
Verbal Aggression. As a criterion measure of verbal aggression, we used the Verbal Hostility
subscale of the Buss–Durkee Hostility Inventory (Buss & Durkee, 1957). Buss (1961) reported
a 5-week test–retest reliability of .72 for this subscale. Supporting construct validity, Verbal
Hostility score is a stronger predictor of hostile content in stories told in response to projective
stimuli (Buss, Fischer, & Simmons, 1962) and of verbal aggression in role-playing responses to
frustrating everyday events (Leibowitz, 1968) compared to the other Buss–Durkee subscales.
Although the original Verbal Hostility subscale had 13 items, we omitted 5 of these (Buss–
144 BRYANT AND SMITH
Durkee items 7, 19, 23, 43, and 63) because of their nearly identical wording with AQ items
to avoid artificially inflating the correlation between the AQ and this criterion measure (cf.
Nichols et al., 1982).
Anger Arousal. As a criterion measure of anger, we chose the Anger Arousal subscale of
the Multidimensional Anger Inventory (MAI; Siegel, 1986). We selected the Anger Arousal
subscale of the MAI as an anger criterion measure because it more closely matches Buss and
Perry’s (1992) conceptual definition of anger as involving ‘‘physiological arousal’’ (p. 457)
compared to other anger scales and because it has been found to be more reliable than the
anger-range, anger-in, and anger-out MAI subscales (Siegel, 1986). Siegel (1986) reported
reliability coefficients for this subscale of .83 for a college sample and .82 for a sample of
factory workers. As validity evidence, Siegel (1986) found Anger Arousal scores correlated
significantly with the magnitude and duration subscales of the Harburg Anger-In/Anger-Out
Inventory (Harburg, Erfurt, Hauenstein, Chape, Schull, & Schork, 1973) and with the magni-
tude subscale of the Novaco Anger Inventory (Novaco, 1975). Although the original MAI
Anger Arousal subscale contained eight items, we used only four of these (MAI items 9, 10,
14, and 26) and omitted items with wording that overlapped the Buss–Perry items to avoid
artificially inflating the correlation between the AQ and this criterion measure (cf. Nichols et
al., 1982).
Hostility. As a criterion measure of global hostility, we chose the Cook–Medley Hostility
Scale (Ho; Cook & Medley, 1954). Based on a subset of 50 true–false items from the Minne-
sota Multiphasic Personality Inventory, the Ho is intended to assess primarily cynicism and
paranoid alienation. Scores on this instrument have a 3-year test–retest correlation of .84
(Shekelle, Gale, Ostfeld, & Paul, 1983). As evidence of prospective validity, the scale appears
to be an independent predictor of later coronary disease (Barefoot, Dahlstrom, & Williams,
1983).
Measurement Models Evaluated in This Study
A systematic literature search uncovered four different measurement models for the AQ
currently in use: (a) a unidimensional ‘‘total score’’ model that assumes a single, global Ag-
gression factor explains responses to all 29 AQ items; (b) Buss and Perry’s (1992) original
four-factor model (Physical Aggression, Verbal Aggression, Anger, and Hostility) for the 29-
item AQ; (c) Buss and Perry’s (1992) hierarchical version of this four-factor model that as-
sumes a single second-order Aggression factor underlies the covariation among the four first-
order factors; and (d) a modified four-factor model for 27 AQ items, consisting of Buss and
Perry’s original Physical Aggression, Verbal Aggression, and Anger factors, along with Har-
ris’s (1995) reduced Hostility factor (omitting its sixth and eighth indicators).
We also evaluated the goodness-of-fit of two new factor structures that we developed as
potential measurement models for the AQ. The first of these new models consisted of our
refined version of Buss and Perry’s original four factors for a subset of 12 AQ items. The
second new model was a hierarchical version of this refined four-factor model that assumes
a single-order Aggression factor underlies the covariation among the four first-order factors.
Overview of Analyses
Our analyses addressed four main questions. (1) Do the existing one-, four-, or hierarchical-
factor structures provide an acceptable measurement model for the AQ? To answer this ques-
tion, we used confirmatory factor analysis (CFA) to impose each of these factor models on
three different data sets and to evaluate each model’s goodness-of-fit across samples. (2) If
available models prove inadequate, then what might be a more appropriate measurement model
for the AQ? Here we used principal components analysis to eliminate unreliable AQ items
in order to develop an acceptable measurement model. (3) Are the same measurement models
ARCHITECTURE OF AGGRESSION 145
warranted for males and females? To address this question, we used multigroup CFA to test
hypotheses about the invariance of the refined AQ measurement model with respect to gender.
(4) Is there any evidence for the convergent or discriminant validity of the multiple AQfactors?
Here we used CFA to test hypotheses about the relationships among the AQ factors and the
criterion measures of aggression.
Analysis Strategy
Stage One. The analysis unfolded in four stages, each using CFA via LISREL 8 (Joreskog &
Sorbom, 1996). In Stage One, we began by using the data from Samples 1–3 to examine the
fit of the four, existing measurement models for the AQ. To evaluate each model’s goodness-
of-fit to the data, we used three measures of absolute fit and two measures of relative fit (cf.
Hu & Bentler, 1998). These multiple measures of model fit provide complementary informa-
tion about how well a particular model explains the data and should not be construed as
redundant (cf. Bollen, 1989; McDonald & Marsh, 1990).
As measures of each model’s absolute fit, we used the ratio of chi-square to degrees of
freedom (χ
2
/df; Hoelter, 1983), the goodness-of-fit index (GFI; Joreskog & Sorbom, 1996),
and the root-mean-square error of approximation (RMSEA; Steiger, 1989). Although the first
two of these measures of model fit have limited utility (Hu & Bentler, 1998), we have reported
them in order to maximize the comparability of our results with those of prior researchers
who reported these fit measures for the same models. In judging absolute fit, smaller ratios
of chi-square to degrees of freedom reflect better absolute fit, with ratios near two considered
acceptable (Hoelter, 1983). Analogous to R
2
in multiple regression, GFI reflects the proportion
of available variation–covariation information in the data that the given model explains, with
larger GFI values representing better model fit. Bentler and Bonett (1980) have suggested that
formal measurement models have a GFI Ն .90. RMSEA reflects the size of the residuals that
result when using the model to predict the data, with smaller values indicating better fit. Ac-
cording to Browne and Cudeck (1993), RMSEA of .05 or lower represents ‘‘close fit,’’
RMSEA between .05 and .08 represents ‘‘reasonably close fit,’’ and RMSEA above .10 repre-
sents ‘‘an unacceptable model.’’ We also directly compared the absolute fit of nested models
by contrasting their goodness-of-fit chi-square values and computing the p value associated
with the difference in these nested chi-squares (with accompanying difference in degrees of
freedom).
As measures of each model’s relative fit, we used the comparative fit index (CFI; Bentler,
1990) and the nonnormed fit index (NNFI; Bentler & Bonett, 1980; Tucker & Lewis, 1973).
We have chosen to report these particular measures because they have more desirable psycho-
metric properties than other measures of relative fit (Bentler, 1990; Hu & Bentler, 1998; Marsh,
Balla, & Hau, 1996; McDonald & Marsh, 1990). Each of these two relative fit measures uses
a different formula to contrast the goodness-of-fit chi-square of a given model with that of a
‘‘null’’ model, which assumes sampling error alone explains the covariation among observed
measures (i.e., that there is no common variance among the AQ items). For each relative fit
index, larger values represent better fit, with values of .90 or higher considered acceptable
(Bentler & Bonett, 1980).
Stage Two. Given the poor fit of the unidimensional and multidimensional models, we
sought to develop a better-fitting, refined measurement model in Stage Two of the analysis.
Stage Two consisted of four phases. In the first phase (model refinement), we subjected the
data of Sample 1 to a principal components analysis (PCA) with oblique rotation in order to
explore the structure underlying the AQ. Extracting four factors, we looked for items with
low communalities or multiple loadings across factors to identify questions that are unreliable
indicators of aggression or that reflect more than one dimension of aggression. We also decided
to omit items that are reverse-scored because we wanted the refined measurement model to
use only indicators that reflect the endorsement of aggression rather than the rejection of
146 BRYANT AND SMITH
nonaggression. This latter approach has been useful in refining measurement models for other
constructs, such as affect intensity (Bryant et al., 1996).
In the second phase of Stage Two (model evaluation), we used CFA to impose a parsimoni-
ous confirmatory version of this new measurement model on the data of Samples 1–3, using
the refined subset of AQ items. We also evaluated a hierarchical form of this model in which
a single, second-order factor explained the relationships among the four first-order factors. In
addition, we used multigroup CFA to evaluate the generalizability of these refined measure-
ment models across Samples 1–3. In these multigroup analyses, we contrasted the goodness-
of-fit chi-squares of two nested CFA models: one constraining the magnitudes of the factor
loadings to be equal for all three samples and the other omitting this invariance constraint.
A statistically significant difference in the chi-square values of these two models (∆χ
2
) indi-
cates that the given model yields different factor loadings across samples (cf. Bryant & Baxter,
1997; Bryant & Yarnold, 1995). Given a significant overall structural difference, we used
multigroup CFA with equality constraints to pinpoint the specific items responsible for cross-
sample differences. In accordance with standard practice in the structural equation modeling
literature, all multigroup analyses were performed using covariance matrices as input (cf.
Joreskog & Sorbom, 1996).
In the third phase of Stage Two, for comparison purposes we tested a condensed, alternative
form of Buss and Perry’s (1992) original four-factor that preserved the full range of item
content while reducing the number of indicators per factor. Here we used a ‘‘partial disaggrega-
tion’’ approach (Bagozzi & Edwards, 1998; Bagozzi & Heatherton, 1994; Hull, Lehn, & Ted-
lie, 1991) to reconfigure Buss and Perry’s (1992) original ‘‘total disaggregation’’ model (which
had nine indicators for PA, five indicators for VA, seven indicators for ANG, and eight indica-
tors for HO) in terms of three indicators for each latent variable. This entailed parceling indi-
vidual items into composite indicators for each factor, so as to modify the ‘‘atomistic’’ original
model into a more ‘‘molecular’’ form. As Bagozzi and Heatherton (1994) have noted, CFA
models containing more than about five measures per factor are unlikely to fit the data satisfac-
torily. For this reason, we sought to test the goodness-of-fit of a condensed form of Buss and
Perry’s original four-factor model and to evaluate its cross-sample generalizability.
In the fourth phase of Stage Two (testing gender invariance), we used multigroup CFA to
determine whether the refined measurement models provided an equivalent goodness-of-fit to
the data of males and females in Samples 1 and 2. This involved contrasting the goodness-
of-fit chi-squares of two nested CFA models: one constraining the magnitudes of the factor
loadings to be equal for males and females and the other omitting this invariance constraint.
A statistically significant difference in the chi-square values of these two models (∆χ
2
) indi-
cates that the given model yields different factor loadings for men and women (cf. Bryant &
Baxter, 1997; Bryant & Yarnold, 1995). Given a significant overall structural difference, we
used multigroup CFA with equality constraints to pinpoint the specific items responsible for
gender differences. Again, all multigroup analyses were performed on covariance matrices
(cf. Joreskog & Sorbom, 1996).
Stage Three. In Stage Three of the analysis, we evaluated and compared the convergent
and discriminant validity of the refined and original versions of the four-factor model. Here
we used CFA with equality constraints to determine whether each of the four AQ factors
showed a stronger relationship with the criterion measure to which it is presumed to correspond
than with the other criterion measures.
Stage Four. Having established the cross-sample generalizability of the refined measurement
model, we sought to reconfirm it in the final stage of the analysis using a new ‘‘short form’’
of the AQ, which contains only the refined subset of 12 items. Here we used CFA to impose
the refined measurement model on the data of Samples 4 and 5, who completed this new,
short form of the AQ. We also used multigroup CFA to assess the generalizability of the
refined measurement model across Samples 4 and 5 as well as the model’s invariance with
respect to gender in both of these samples.
ARCHITECTURE OF AGGRESSION 147
RESULTS AND DISCUSSION
Stage One: Assessing Current Measurement Models
We used CFA first to evaluate the goodness-of-fit of the four alternative
measurement models currently in the literature, using the AQ data of Sam-
ples 1–3. Which model, if any, provides the most reasonable representation
of responses to this instrument? Table 1 presents the results of these analyses.
Across all three samples, the explanatory power of existing measurement
models fell short of accepted standards. First, a single ‘‘AQ total score’’
provided an inadequate measurement model for Samples 1–3, leaving too
much common variance unexplained in both an absolute (GFIs ϭ.65 Ϫ.70)
and relative (fit indices ϭ .59 Ϫ .66) sense. For all three samples, this one-
factor model also had a relatively high ratio of chi-square to degrees of free-
dom (χ
2
/dfs ϭ3.4 Ϫ4.2) and an unacceptably large, root-mean-square error
of approximation (RMSEAs ϭ .098 Ϫ .109). These findings support the
AQ’s multidimensionality.
Yet, none of the multidimensional models currently available provides an
acceptable measurement model for the AQ. Both Buss and Perry’s (1992)
original four-factor model and its hierarchical counterpart fit the data of all
three samples better than the unidimensional model, all ∆χ
2
s(6) Ͼ 379.2, ps
Ͻ.0001. However, for all three samples, each of these multifactor structures
fails to achieve sufficient goodness-of-fit to be an acceptable measurement
model in both absolute (GFIs ϭ.76 Ϫ .81) and relative (fit indices ϭ .76 Ϫ
.82) terms. Furthermore, the chi-square to degrees of freedom ratios for these
models (χ
2
/dfs ϭ2.4 Ϫ2.8) show ‘‘a poor fit’’ (Buss & Perry, 1992, p. 454),
and their RMSEAs show room for improvement (RMSEAs ϭ .072 Ϫ .084).
Evidently, current measurement models of the AQ operationalize underlying
subtraits of aggression in ways that do not correspond closely enough to the
conceptual framework that Buss and Perry (1995) intended. This conclusion
is consistent with those of previous researchers (e.g., Archer, Kilpatrick, &
Bramwell, 1995; Harris, 1995; Williams et al., 1996) who have noted the
inadequacy of existing factor models for the AQ.
Imposed on the data of Samples 1–3, the four-factor model containing
Harris’s (1995) condensed Hostility factor fared little better (see Table 1).
Although this modified model was a significant improvement in fit over the
original four-factor model for all three samples, all ∆χ
2
(53)s Ͼ 143.6, p Ͻ
.0001, it nevertheless fell short of accepted standards for a formal measure-
ment in both an absolute (GFIs ϭ .78 Ϫ .83) and relative (fit indices ϭ
.79 Ϫ .83) sense. Taken as a whole, this evidence underscores the need for
a better fitting measurement model for the AQ.
Stage Two: Developing a Better Fitting Measurement Model
What might a more appropriate measurement model look like, and how
should we best go about developing it? In answering these questions, we
148 BRYANT AND SMITH
T
A
B
L
E
1
G
o
o
d
n
e
s
s
-
o
f
-
F
i
t
S
t
a
t
i
s
t
i
c
s
f
o
r
V
a
r
i
o
u
s
M
e
a
s
u
r
e
m
e
n
t
M
o
d
e
l
s
o
f
t
h
e
A
Q
I
m
p
o
s
e
d
o
n
S
a
m
p
l
e
s
1

3
R
e
l
a
t
i
v
e
A
b
s
o
l
u
t
e

t
m
e
a
s
u
r
e
s

t
m
e
a
s
u
r
e
s
M
o
d
e
l
N
o
.
i
t
e
m
s
S
a
m
p
l
e
χ
2
d
f
χ
2
/
d
f
G
F
I
R
M
S
E
A
C
F
I
N
N
F
I
O
n
e
-
f
a
c
t
o
r
2
9
1
1
5
6
7
.
9
3
7
7
4
.
2
.
7
0
.
1
0
2
.
6
6
.
6
4
(
t
o
t
a
l
s
c
o
r
e
)
2
1
2
6
7
.
8
3
7
7
3
.
4
.
6
5
.
1
0
9
.
6
2
.
5
9
3
1
4
6
9
.
1
3
7
7
3
.
9
.
6
8
.
0
9
8
.
6
6
.
6
3
B
u
s
s
&
P
e
r
r
y

s
2
9
1
1
0
4
2
.
8
3
7
1
2
.
8
.
8
1
.
0
7
7
.
8
1
.
7
9
f
o
u
r
f
a
c
t
o
r
s
:
2
8
8
6
.
4
3
7
1
2
.
4
.
7
6
.
0
8
4
.
7
8
.
7
6
P
A
,
V
A
,
A
N
G
,
H
O
3
9
5
0
.
3
3
7
1
2
.
6
.
8
1
.
0
7
2
.
8
2
.
8
0
B
u
s
s
&
P
e
r
r
y

s
2
9
1
1
0
4
6
.
4
3
7
3
2
.
8
.
8
1
.
0
7
7
.
8
1
.
7
9
h
i
e
r
a
r
c
h
i
c
a
l
m
o
d
e
l
:
2
8
8
8
.
5
3
7
3
2
.
4
.
7
6
.
0
8
3
.
7
8
.
7
6
o
n
e
s
e
c
o
n
d
-
o
r
d
e
r
f
a
c
t
o
r
3
9
6
9
.
6
3
7
3
2
.
6
.
8
1
.
0
7
2
.
8
1
.
8
0
B
u
s
s
&
P
e
r
r
y

s
2
7
1
8
8
1
.
9
3
1
8
2
.
9
.
8
2
.
0
7
6
.
8
3
.
8
2
P
A
,
V
A
,
A
N
G
,
&
2
7
3
4
.
2
3
1
8
2
.
3
.
7
8
.
0
8
1
.
8
1
.
7
9
H
a
r
r
i
s

s
H
O
f
a
c
t
o
r
3
8
0
6
.
6
3
1
8
2
.
5
.
8
3
.
0
7
1
.
8
3
.
8
1
F
o
u
r
r
e

n
e
d
f
a
c
t
o
r
s
:
1
2
1
1
0
5
.
7
4
8
2
.
2
.
9
4
.
0
6
3
.
9
6
.
9
4
P
A
,
V
A
,
A
N
G
,
H
O
2
9
2
.
4
4
8
1
.
9
.
9
3
.
0
6
8
.
9
5
.
9
3
3
1
2
1
.
7
4
8
2
.
5
.
9
4
.
0
7
1
.
9
1
.
8
7
R
e

n
e
d
h
i
e
r
a
r
c
h
i
c
a
l
1
2
1
1
0
8
.
5
5
0
2
.
2
.
9
4
.
0
6
2
.
9
6
.
9
4
m
o
d
e
l
:
2
9
4
.
4
5
0
1
.
9
.
9
3
.
0
6
7
.
9
5
.
9
3
o
n
e
s
e
c
o
n
d
-
o
r
d
e
r
f
a
c
t
o
r
3
1
3
3
.
6
5
0
2
.
7
.
9
3
.
0
7
4
.
9
0
.
8
6
N
o
t
e
.
P
A
ϭ
P
h
y
s
i
c
a
l
A
g
g
r
e
s
s
i
o
n
;
V
A
ϭ
V
e
r
b
a
l
A
g
g
r
e
s
s
i
o
n
;
A
N
G
ϭ
A
n
g
e
r
;
H
O
ϭ
H
o
s
t
i
l
i
t
y
;
G
F
I
ϭ
g
o
o
d
n
e
s
s
-
o
f
-

t
i
n
d
e
x
(
J
o
r
e
s
k
o
g
&
S
o
r
b
o
m
,
1
9
9
6
)
;
R
M
S
E
A
ϭ
r
o
o
t
-
m
e
a
n
-
s
q
u
a
r
e
e
r
r
o
r
o
f
a
p
p
r
o
x
i
m
a
t
i
o
n
(
S
t
e
i
g
e
r
,
1
9
8
9
)
;
C
F
I
ϭ
c
o
m
p
a
r
a
t
i
v
e

t
i
n
d
e
x
(
B
e
n
t
l
e
r
,
1
9
9
0
)
;
N
N
F
I
ϭ
n
o
n
n
o
r
m
e
d

t
i
n
d
e
x
(
B
e
n
t
l
e
r
&
B
o
n
e
t
t
,
1
9
8
0
)
.
S
a
m
p
l
e
1
ϭ
3
0
7
A
m
e
r
i
c
a
n
u
n
d
e
r
g
r
a
d
u
a
t
e
s
;
S
a
m
p
l
e
2
ϭ
2
0
0
B
r
i
t
i
s
h
u
n
d
e
r
g
r
a
d
u
a
t
e
s
(
A
r
c
h
e
r
,
H
o
l
l
o
w
a
y
,
&
M
c
L
o
u
g
h
l
i
n
,
1
9
9
5
)
;
S
a
m
p
l
e
3
ϭ
3
0
6
C
a
n
a
d
i
a
n
u
n
d
e
r
g
r
a
d
u
a
t
e
s
(
H
a
r
r
i
s
,
1
9
9
5
)
.
T
h
e
a
b
o
v
e
g
o
o
d
n
e
s
s
-
o
f
-

t
s
t
a
t
i
s
t
i
c
s
a
r
e
f
r
o
m
a
n
a
l
y
s
e
s
c
o
n
d
u
c
t
e
d
v
i
a
L
I
S
R
E
L
8
(
J
o
r
e
s
k
o
g
&
S
o
r
b
o
m
,
1
9
9
6
)
.
ARCHITECTURE OF AGGRESSION 149
sought first to preserve the solid theoretical foundation underlying Buss and
Perry’s (1992) original four-factor model, while at the same time sharpening
its measurement focus. Accordingly, we set out not only to maintain the
conceptual definitions of the original factors, but also to improve the four-
factor model’s goodness-of-fit by eliminating AQ items that were relatively
unreliable indicators of the dimensions they were intended to reflect.
Model refinement. With these goals in mind, we inspected the results from
principal components analysis (PCA) of Sample 1 data and decided to ex-
clude AQ items according to three criteria. First, we eliminated items with
low loadings (λ Ͻ .40) in order to increase the proportion of variance that
factors explained in their constituent indicators. Second, to enhance the con-
ceptual clarity of the model, we excluded items that loaded at least moder-
ately (λ ϭ .40) on two or more scales, based on CFA modification indices
and PCA results. Third, to improve conceptual precision, we omitted items
that did not reflect the direct endorsement of aggressive traits. The AQ in-
cludes two reverse-scored items—a Physical Aggression item (‘‘I can think
of no good reason for ever hitting a person’’) and an Anger item (‘‘I am an
even-tempered person’’)—which entail the rejection of nonaggressive traits
rather than the acceptance of aggressive characteristics. These three exclu-
sion criteria yielded a refined 12-item, four-factor model that reflects the
same underlying constructs (Physical Aggression, Verbal Aggression,
Anger, and Hostility) as Buss and Perry’s (1992) original model, but with
an equal number of items for each factor. Table 2 presents the items compris-
ing both the original and refined versions of the four-factor model of the
AQ.
Model evaluation. How well does this refined four-factor model explain
responses to the AQ? To address this question, we used CFA to impose the
refined measurement model on the data of Samples 1–3. In doing this, we
have used Sample 1 to modify the model post hoc and used Samples 2 and
3 to cross-validate the model a priori. This strategy enabled us to assess the
degree to which model respecifications based on Sample 1 generalized across
independent samples, so as to avoid being misled by the unique characteris-
tics of a single sample (cf. MacCallum, 1986; MacCallum, Roznowski, &
Necowitz, 1992).
Table 1 also presents the results of these analyses. As seen in this table,
across all three samples, the refined four-factor model explains an acceptable
proportion of common variance in both absolute (GFIs ϭ .93 Ϫ .94) and
relative (fit indices ϭ .87 Ϫ .96) terms. Although its ratio of chi-square to
degrees of freedom shows cross-sample inconsistency (χ
2
/dfs ϭ 1.9 Ϫ 2.5),
the refined model’s RMSEA reflects reasonably close fit across all three sam-
ples (RMSEAs ϭ .063 Ϫ .071). Considered together, these findings suggest
that the modified four-factor model is an appropriate measurement model
for the AQ.
150 BRYANT AND SMITH
TABLE 2
Items Constituting the Original and Refined Measurement Models of the AQ
Factor Constituent items
Physical Aggression 1. Once in a while I can’t control the urge to hit another person.
2. Given enough provocation, I may hit another person.
3. If somebody hits me, I hit back.
4. I get into fights a little more than the average person.
5. If I have to resort to violence to protect my rights, I will.
6. There are people who pushed me so far that we came to
blows.
7. I can think of no good reason for ever hitting a person.
[reverse-scored]
8. I have threatened people I know.
9. I have become so mad that I have broken things.
Verbal Aggression 10. I tell my friends openly when I disagree with them.
11. I often find myself disagreeing with people.
12. When people annoy me, I may tell them what I think of them.
13. I can’t help getting into arguments when people disagree
with me.
14. My friends say that I’m somewhat argumentative.
Anger 15. I flare up quickly but get over it quickly.
16. When frustrated, I let my irritation show.
17. I sometimes feel like a powder keg ready to explode.
18. I am an even-tempered person. [reverse-scored]
19. Some of my friends think I’m a hothead.
20. Sometimes I fly off the handle for no good reason.
21. I have trouble controlling my temper.
Hostility 22. I am sometimes eaten up with jealousy.
23. At times I feel I have gotten a raw deal out of life.
24. Other people always seem to get the breaks.
25. I wonder why sometimes I feel so bitter about things.
26. I know that ‘‘friends’’ talk about me behind my back.
27. I am suspicious of overly friendly strangers.
28. I sometimes feel that people are laughing at me behind my
back.
29. When people are especially nice, I wonder what they want.
Note: The 29 items constituting the original four-factor model for the Aggression Question-
naire (AQ) are listed in the order presented by Buss and Perry (1992, Table 1, p. 454). Buss
and Perry (1992) instructed researchers to randomly order the above items when administering
the AQ. The order of the 29 AQ items for Sample 1 was 26, 15, 11, 10, 16, 27, 2, 4, 29, 12,
21, 13, 14, 8, 5, 22, 28, 7, 20, 6, 25, 19, 23, 3, 24, 1, 9, 17, and 18. Items in bold comprise
the refined four-factor measurement model. The randomized order of these questions in the
12-item short form of the AQ is 11, 23, 8, 25, 21, 14, 15, 2, 13, 24, 6, and 20. From ‘‘The
Aggression Questionnaire,’’ by A. H. Buss and M. Perry, 1992, Journal of Personality and
Social Psychology, 63, p. 454 (Table 1). Copyright by the American Psychological Associa-
tion. Adapted with permission.
ARCHITECTURE OF AGGRESSION 151
Results converge on an identical conclusion concerning the hierarchical
version of this refined model. Specifically, a model specifying a single, over-
arching second-order aggression trait that explains the covariation among
the four first-order factors fit the data of all three samples reasonably well,
both absolutely (GFIs ϭ .93 Ϫ .94) and relatively (fit indices ϭ .86 Ϫ .96).
This hierarchical model also showed chi-square to degrees of freedom ratios

2
/dfs ϭ 1.9 Ϫ 2.7) and RMSEAs (ϭ .062 Ϫ .074) comparable to those
of the refined four-factor model across the three samples. Thus, both the
refined four-factor model and its hierarchical counterpart represent accept-
able measurement models for the AQ.
Testing cross-sample generalizability. Having identified an appropriate
measurement structure, we next used multigroup CFA to evaluate the cross-
sample generalizability of the four-factor model more systematically. We
considered first the question of whether the refined four-factor model pro-
duces the same factor loadings across the three samples. In other words, do
Physical Aggression, Verbal Aggression, Anger, and Hostility have the same
meanings for American (Sample 1), British (Sample 2), and Canadian (Sam-
ple 3) undergraduates?
Table 3 presents the loadings for the four-factor model imposed on the
data of each sample. An initial omnibus test revealed that the magnitudes
of these factor loadings varied across samples, ∆χ
2
(16, n ϭ 813) ϭ 29.3,
p Ͻ .022. Following up the omnibus test, only Samples 1 and 3 showed
significant differences in factor loadings, ∆χ
2
(8, n ϭ613) ϭ24.3, p Ͻ.003,
whereas loadings were equivalent for Samples 1 and 2, ∆χ
2
(8, n ϭ 507) ϭ
11.1, p Ͼ .19; and Samples 2 and 3, ∆χ
2
(8, n ϭ 506) ϭ 7.7, p Ͼ .46. Addi-
tional multigroup CFAs disclosed that only one factor loading actually dif-
fered significantly for Samples 1 and 3—the loading for Anger item 1 (‘‘I
flare up quickly but get over it quickly’’) was stronger for the Canadian
sample, ∆χ
2
(1, n ϭ 613) ϭ 13.8, p Ͻ .0003—and all other loadings for the
four-factor model were statistically comparable, ∆χ
2
(7, n ϭ 613) ϭ 12.5,
p Ͼ .09. Supporting this conclusion, a model that constrains (a) all factor
loadings except Anger item 1 to be equal across Samples 1–3 and (b) Anger
item 1 to load equally for Samples 1 and 2 but not for Sample 3 fit the data
of the three samples no worse than a multigroup model with no equality
constraints, ∆χ
2
(15, n ϭ 813) ϭ 19.5, p Ͼ .19. These results indicate that
the loadings of the four-factor model are largely (11/12 ϭ 92%) invariant
across the three samples; they also suggest that the meaning of aggression,
as defined by the refined measurement model, holds across culture.
The same cannot be said of Buss and Perry’s (1992) original four-factor
model, which produced nonequivalent loadings for all three samples, ∆χ
2
(50,
n ϭ813) ϭ554.8, p Ͻ.0001. Indeed, the original four-factor model yielded
strong differences in loadings when comparing Samples 1 and 2, ∆χ
2
(25,
n ϭ 507) ϭ 67.6, p Ͻ .0001; Samples 2 and 3, ∆χ
2
(25, n ϭ 506) ϭ 634.4,
152 BRYANT AND SMITH
T
A
B
L
E
3
C
F
A
F
a
c
t
o
r
L
o
a
d
i
n
g
s
f
o
r
t
h
e
R
e

n
e
d
1
2
-
I
t
e
m
,
F
o
u
r
-
F
a
c
t
o
r
M
e
a
s
u
r
e
m
e
n
t
M
o
d
e
l
o
f
t
h
e
A
Q
P
A
V
A
A
N
G
H
O
s
a
m
p
l
e
s
a
m
p
l
e
s
a
m
p
l
e
s
a
m
p
l
e
A
Q
i
t
e
m
s
1
2
3
1
2
3
1
2
3
1
2
3
2
.
G
i
v
e
n
e
n
o
u
g
h
p
r
o
v
o
c
a
t
i
o
n
,
I
m
a
y
h
i
t
a
n
o
t
h
e
r
p
e
r
s
o
n
.
7
6
7
0
5
8
6
.
T
h
e
r
e
a
r
e
p
e
o
p
l
e
w
h
o
p
u
s
h
e
d
m
e
s
o
f
a
r
t
h
a
t
w
e
c
a
m
e
t
o
b
l
o
w
s
.
7
2
7
3
6
5
8
.
I
h
a
v
e
t
h
r
e
a
t
e
n
e
d
p
e
o
p
l
e
I
k
n
o
w
.
8
0
8
2
6
8
1
1
.
I
o
f
t
e
n

n
d
m
y
s
e
l
f
d
i
s
a
g
r
e
e
i
n
g
w
i
t
h
p
e
o
p
l
e
.
8
0
7
5
7
0
1
3
.
I
c
a
n

t
h
e
l
p
g
e
t
t
i
n
g
i
n
t
o
a
r
g
u
m
e
n
t
s
w
h
e
n
p
e
o
p
l
e
d
i
s
a
g
r
e
e
w
i
t
h
m
e
.
8
2
7
1
6
8
1
4
.
M
y
f
r
i
e
n
d
s
s
a
y
t
h
a
t
I

m
s
o
m
e
w
h
a
t
a
r
g
u
m
e
n
t
a
t
i
v
e
.
5
8
6
1
7
6
1
5
.
I

a
r
e
u
p
q
u
i
c
k
l
y
b
u
t
g
e
t
o
v
e
r
i
t
q
u
i
c
k
l
y
.
5
0
6
2
6
9
2
0
.
S
o
m
e
t
i
m
e
s
I

y
o
f
f
t
h
e
h
a
n
d
l
e
f
o
r
n
o
g
o
o
d
r
e
a
s
o
n
8
1
8
3
5
7
2
1
.
I
h
a
v
e
t
r
o
u
b
l
e
c
o
u
n
t
r
o
l
l
i
n
g
m
y
t
e
m
p
e
r
.
7
1
7
1
3
4
2
3
.
A
t
t
i
m
e
s
I
f
e
e
l
I
h
a
v
e
g
o
t
t
e
n
a
r
a
w
d
e
a
l
o
u
t
o
f
l
i
f
e
.
6
5
7
6
4
5
2
4
.
O
t
h
e
r
p
e
o
p
l
e
a
l
w
a
y
s
s
e
e
m
t
o
g
e
t
t
h
e
b
r
e
a
k
s
.
7
7
7
5
6
4
2
5
.
I
w
o
n
d
e
r
w
h
y
s
o
m
e
t
i
m
e
s
I
f
e
e
l
s
o
b
i
t
t
e
r
a
b
o
u
t
t
h
i
n
g
s
.
6
8
6
8
5
2
N
o
t
e
.
P
A
ϭ
P
h
y
s
i
c
a
l
A
g
g
r
e
s
s
i
o
n
;
V
A
ϭ
V
e
r
b
a
l
A
g
g
r
e
s
s
i
o
n
;
A
N
G
ϭ
A
n
g
e
r
;
H
O
ϭ
H
o
s
t
i
l
i
t
y
.
D
e
c
i
m
a
l
p
o
i
n
t
s
a
r
e
o
m
i
t
t
e
d
.
I
t
e
m
n
u
m
b
e
r
s
i
n
p
a
r
e
n
t
h
e
s
e
s
r
e
f
e
r
t
o
t
h
e
o
r
i
g
i
n
a
l
o
r
d
e
r
i
n
g
o
f
i
t
e
m
s
i
n
B
u
s
s
a
n
d
P
e
r
r
y

s
(
1
9
9
2
,
p
.
4
5
4
)
T
a
b
l
e
1
.
B
l
a
n
k
l
o
a
d
i
n
g
s
w
e
r
e

x
e
d
a
t
z
e
r
o
.
T
h
e
l
o
a
d
i
n
g
s
o
f
i
t
e
m
s
2
,
1
4
,
2
0
,
a
n
d
2
3
w
e
r
e

x
e
d
a
t
u
n
s
t
a
n
d
a
r
d
i
z
e
d
v
a
l
u
e
s
o
f
1
.
0
t
o
s
c
a
l
e
t
h
e
l
a
t
e
n
t
v
a
r
i
a
b
l
e
s
i
n
m
u
l
t
i
g
r
o
u
p
c
o
n

r
m
a
t
o
r
y
a
n
a
l
y
s
e
s
.
ARCHITECTURE OF AGGRESSION 153
p Ͻ .0001; and Samples 1 and 3, ∆χ
2
(25, n ϭ 613) ϭ 126.9, p Ͻ .0001.
Clearly, the refined measurement model is superior to the original in terms
of both its goodness-of-fit and its cross-cultural generalizability.
Comparing the refined hierarchical model across samples, the second-
order Aggression factor had the same relationships with the four first-order
factors in Samples 1 and 2, χ
2
(3, n ϭ507) ϭ0.6, p Ͼ.89; but the Aggression
‘‘super factor’’ had less to do with Anger in Canadian Sample 3 compared
to American Sample 1, χ
2
(1, n ϭ613) ϭ25.5, p Ͻ.0001; and British Sample
2, χ
2
(1, n ϭ 506) ϭ 24.6, p Ͻ .0001 (see Table 4). These results are consis-
tent with the earlier finding that the four-factor model fits the data of Sample
3 better than the hierarchical model.
Comparing the original hierarchical model across samples, the second-
order Aggression factor had the same relationships with the four first-order
factors in Samples 1 and 2, χ
2
(3, n ϭ507) ϭ1.9, p Ͼ.59; but the Aggression
‘‘super factor’’ had less to do with Anger in Canadian Sample 3 compared
to American Sample 1, χ
2
(1, n ϭ 613) ϭ 22.6, p Ͻ .00005; and British
Sample 2, χ
2
(1, n ϭ 506) ϭ 28.8, p Ͻ .00001 (see Table 4). Unlike the
refined hierarchical model, however, second-order Aggression also had more
to do with Hostility in Canadian Sample 3 compared to the American Sample
1, χ
2
(1, n ϭ613) ϭ34.0, p Ͻ .00001; and British Sample 2, χ
2
(1, n ϭ 506)
ϭ 16.6, p Ͻ.0009 (see Table 4). Thus, the refined second-order CFA model
showed stronger cross-cultural generalizability than did the original second-
order CFA model.
We also addressed the question of whether the four factors comprising
the refined model interrelate in the same ways for the American, British, and
Canadian samples. Table 5 presents the reliabilities and factor intercorrela-
tions for this model imposed on the data of Samples 1–3. Using the model
with partially invariant loadings as a baseline (cf. Byrne, Shavelson, & Mu-
then, 1989), the refined four-factor model produced different factor correla-
tions for the three samples, ∆χ
2
(12, n ϭ 813) ϭ 65.3, p Ͻ .0001. Although
Samples 1 and 2 had equivalent factor intercorrelations, ∆χ
2
(6, n ϭ 507) ϭ
3.1, p Ͼ .79; Sample 3 had different factor intercorrelations compared to
Sample 1, ∆χ
2
(6, n ϭ 613) ϭ 53.1, p Ͻ .0001; and Sample 2, ∆χ
2
(6, n ϭ
506) ϭ46.3, p Ͻ.0001. Thus, the four refined AQ factors interrelated differ-
ently among the Canadian sample than among the American and British
samples.
Although the correlation between Physical and Verbal Aggression was
the same in all three samples, ∆χ
2
(2) ϭ 4.2, p Ͼ .12, Physical Aggression
correlated more strongly with Anger in the Canadian sample than in the other
two samples, both ∆χ
2
(1)s Ͼ5.1, ps Ͻ .025. For the Canadian sample, Ver-
bal Aggression also correlated more strongly with Anger, both ∆χ
2
(1)s Ͼ
14.0, ps Ͻ .002; but Hostility correlated less strongly with Physical Ag-
gression, both ∆χ
2
(1)s Ͼ 9.2, ps Ͻ .0025, with Verbal Aggression, both
154 BRYANT AND SMITH
T
A
B
L
E
4
S
e
c
o
n
d
-
O
r
d
e
r
F
a
c
t
o
r
L
o
a
d
i
n
g
s
a
n
d
R
e
s
i
d
u
a
l
V
a
r
i
a
n
c
e
s
o
f
t
h
e
F
i
r
s
t
-
O
r
d
e
r
F
a
c
t
o
r
s
f
o
r
t
h
e
O
r
i
g
i
n
a
l
a
n
d
R
e

n
e
d
H
i
e
r
a
r
c
h
i
c
a
l
C
F
A
M
o
d
e
l
I
m
p
o
s
e
d
o
n
t
h
e
A
Q
D
a
t
a
o
f
S
a
m
p
l
e
s
1

3
S
a
m
p
l
e
1
S
a
m
p
l
e
2
S
a
m
p
l
e
3
A
Q
f
a
c
t
o
r
s
M
o
d
e
l
P
A
V
A
A
N
G
H
O
P
A
V
A
A
N
G
H
O
P
A
V
A
A
N
G
H
O
F
a
c
t
o
r
l
o
a
d
i
n
g
O
r
i
g
i
n
a
l
1
.
0
0
*
.
9
9
1
.
0
1
.
5
4
1
.
0
0
*
1
.
1
8
1
.
3
3
.
6
3
1
.
0
0
*
.
9
3
.
4
2
1
.
4
4
(
γ
)
R
e

n
e
d
1
.
0
0
*
.
8
0
1
.
2
1
.
7
7
1
.
0
0
*
.
7
5
1
.
2
3
.
8
1
1
.
0
0
*
.
9
6
.
4
7
1
.
0
1
S
t
a
n
d
a
r
d
e
r
r
o
r
o
f
O
r
i
g
i
n
a
l

.
1
0
.
1
1
.
0
7

.
1
6
.
1
7
.
1
1

.
1
2
.
0
8
.
1
7
f
a
c
t
o
r
l
o
a
d
i
n
g
R
e

n
e
d

.
1
2
.
1
6
.
1
2

.
1
3
.
2
0
.
1
6

.
1
6
.
1
1
.
1
8
S
t
a
n
d
a
r
d
i
z
e
d
O
r
i
g
i
n
a
l
.
7
6
.
8
0
.
9
9
.
6
4
.
7
6
.
8
6
.
9
5
.
6
1
.
7
5
.
6
5
.
6
3
.
9
9
f
a
c
t
o
r
l
o
a
d
i
n
g
R
e

n
e
d
.
7
1
.
8
3
.
9
0
.
6
6
.
7
7
.
8
4
.
8
6
.
5
4
.
7
1
.
6
2
.
7
0
.
9
9
S
q
u
a
r
e
d
m
u
l
t
i
p
l
e
O
r
i
g
i
n
a
l
.
5
7
.
6
4
.
9
9
.
4
1
.
5
8
.
7
3
.
9
0
.
3
7
.
5
6
.
4
3
.
4
0
.
9
9
c
o
r
r
e
l
a
t
i
o
n
R
e

n
e
d
.
5
0
.
6
8
.
8
1
.
4
3
.
5
9
.
7
0
.
7
4
.
2
9
.
5
1
.
3
8
.
4
8
.
9
9
R
e
s
i
d
u
a
l
v
a
r
i
a
n
c
e
O
r
i
g
i
n
a
l
.
5
5
.
4
0
.
0
1
.
3
1
.
3
5
.
2
5
.
1
0
.
3
3
.
3
3
.
4
8
.
1
1
.
0
1
(
Ψ
)
R
e

n
e
d
.
3
8
.
1
1
.
1
3
.
3
0
.
3
0
.
1
0
.
2
2
.
7
0
.
3
2
.
5
1
.
0
8
.
0
1
S
t
a
n
d
a
r
d
e
r
r
o
r
o
f
O
r
i
g
i
n
a
l
.
0
8
.
0
7
.
0
4
.
0
6
.
0
7
.
0
8
.
0
6
.
0
7
.
0
6
.
0
8
.
0
3
.
0
6
r
e
s
i
d
u
a
l
v
a
r
i
a
n
c
e
R
e

n
e
d
.
0
7
.
1
1
.
0
5
.
0
6
.
0
8
.
0
4
.
0
8
.
1
4
.
0
8
.
0
9
.
0
3
.
0
4
S
t
a
n
d
a
r
d
i
z
e
d
O
r
i
g
i
n
a
l
.
4
3
.
3
6
.
0
1
.
5
9
.
4
2
.
2
7
.
1
0
.
6
3
.
4
4
.
5
7
.
6
0
.
0
1
r
e
s
i
d
u
a
l
v
a
r
i
a
n
c
e
R
e

n
e
d
.
5
0
.
3
2
.
1
9
.
5
7
.
4
1
.
3
0
.
2
6
.
7
1
.
4
9
.
6
2
.
5
2
.
0
1
N
o
t
e
:
P
A
ϭ
P
h
y
s
i
c
a
l
A
g
g
r
e
s
s
i
o
n
;
V
A
ϭ
V
e
r
b
a
l
A
g
g
r
e
s
s
i
o
n
;
A
N
G
ϭ
A
n
g
e
r
;
H
O
ϭ
H
o
s
t
i
l
i
t
y
.
S
a
m
p
l
e
1
ϭ
3
0
7
A
m
e
r
i
c
a
n
u
n
d
e
r
g
r
a
d
u
a
t
e
s
;
S
a
m
p
l
e
2
ϭ
2
0
0
B
r
i
t
i
s
h
u
n
d
e
r
g
r
a
d
u
a
t
e
s
(
A
r
c
h
e
r
,
H
o
l
l
o
w
a
y
,
&
M
c
L
o
u
g
h
l
i
n
,
1
9
9
5
)
;
S
a
m
p
l
e
3
ϭ
3
0
6
C
a
n
a
d
i
a
n
u
n
d
e
r
g
r
a
d
u
a
t
e
s
(
H
a
r
r
i
s
,
1
9
9
5
)
.
T
a
b
l
e
d
a
r
e
r
e
s
u
l
t
s
f
r
o
m
s
e
c
o
n
d
-
o
r
d
e
r
C
F
A
m
o
d
e
l
s
i
m
p
o
s
e
d
s
e
p
a
r
a
t
e
l
y
o
n
t
h
e
d
a
t
a
o
f
e
a
c
h
s
a
m
p
l
e
.
S
e
c
o
n
d
-
o
r
d
e
r
f
a
c
t
o
r
l
o
a
d
i
n
g
s
(
γ
s
)
a
r
e
u
n
s
t
a
n
d
a
r
d
-
i
z
e
d
r
e
g
r
e
s
s
i
o
n
c
o
e
f

c
i
e
n
t
s
r
e
p
r
e
s
e
n
t
i
n
g
t
h
e
a
m
o
u
n
t
o
f
c
h
a
n
g
e
i
n

r
s
t
-
o
r
d
e
r
f
a
c
t
o
r
s
c
o
r
e
s
r
e
s
u
l
t
i
n
g
f
r
o
m
a
o
n
e
-
u
n
i
t
c
h
a
n
g
e
i
n
t
h
e
s
e
c
o
n
d
-
o
r
d
e
r
l
a
t
e
n
t
v
a
r
i
a
b
l
e
.
S
t
a
n
d
a
r
d
i
z
e
d
s
e
c
o
n
d
-
o
r
d
e
r
f
a
c
t
o
r
l
o
a
d
i
n
g
s
a
r
e
s
t
a
n
d
a
r
d
i
z
e
d
r
e
g
r
e
s
s
i
o
n
c
o
e
f

c
i
e
n
t
s
r
e
p
r
e
s
e
n
t
i
n
g
t
h
e
c
h
a
n
g
e
i
n
s
t
a
n
d
a
r
d
d
e
v
i
a
t
i
o
n
s
i
n

r
s
t
-
o
r
d
e
r
f
a
c
t
o
r
s
c
o
r
e
s
r
e
s
u
l
t
i
n
g
f
r
o
m
a
1
-
s
t
a
n
d
a
r
d
-
d
e
v
i
a
t
i
o
n
c
h
a
n
g
e
i
n
t
h
e
s
e
c
o
n
d
-
o
r
d
e
r
f
a
c
t
o
r
.
S
q
u
a
r
e
d
m
u
l
t
i
p
l
e
c
o
r
r
e
l
a
t
i
o
n
s
r
e
p
r
e
s
e
n
t
t
h
e
p
r
o
p
o
r
t
i
o
n
o
f
t
o
t
a
l
v
a
r
i
a
n
c
e
i
n
e
a
c
h

r
s
t
-
o
r
d
e
r
f
a
c
t
o
r
t
h
a
t
i
s
a
s
s
o
c
i
a
t
e
d
w
i
t
h
t
h
e
s
e
c
o
n
d
-
o
r
d
e
r
f
a
c
t
o
r
,
a
n
a
l
o
g
o
u
s
t
o
R
2
i
n
m
u
l
t
i
p
l
e
r
e
g
r
e
s
s
i
o
n
.
R
e
s
i
d
u
a
l
v
a
r
i
a
n
c
e
s
(
ψ
s
)
r
e
p
r
e
s
e
n
t
t
h
e
a
m
o
u
n
t
o
f
v
a
r
i
a
n
c
e
i
n

r
s
t
-
o
r
d
e
r
f
a
c
t
o
r
s
t
h
a
t
i
s
u
n
e
x
p
l
a
i
n
e
d
b
y
t
h
e
s
e
c
o
n
d
-
o
r
d
e
r
f
a
c
t
o
r
.
S
t
a
n
d
a
r
d
i
z
e
d
r
e
s
i
d
u
a
l
v
a
r
i
a
n
c
e
s
r
e
p
r
e
s
e
n
t
t
h
e
p
r
o
p
o
r
t
i
o
n
o
f
t
o
t
a
l
v
a
r
i
a
n
c
e
i
n
e
a
c
h

r
s
t
-
o
r
d
e
r
f
a
c
t
o
r
t
h
a
t
i
s
u
n
r
e
l
a
t
e
d
t
o
t
h
e
s
e
c
o
n
d
-
o
r
d
e
r
f
a
c
t
o
r
a
n
d
t
h
a
t
i
s
s
p
e
c
i

c
t
o
t
h
a
t

r
s
t
-
o
r
d
e
r
f
a
c
t
o
r
.
S
t
a
n
d
a
r
d
e
r
r
o
r
s
o
f
s
e
c
o
n
d
-
o
r
d
e
r
f
a
c
t
o
r
l
o
a
d
i
n
g
s
(
a
n
d
o
f
r
e
s
i
d
u
a
l
v
a
r
i
a
n
c
e
s
)
r
e
p
r
e
s
e
n
t
t
h
e
s
t
a
n
d
a
r
d
d
e
v
i
a
t
i
o
n
o
f
c
h
a
n
g
e
s
i
n
l
o
a
d
i
n
g
s
(
a
n
d
i
n
r
e
s
i
d
u
a
l
v
a
r
i
a
n
c
e
s
)
t
h
a
t
w
o
u
l
d
b
e
e
x
p
e
c
t
e
d
t
o
o
c
c
u
r
f
r
o
m
s
a
m
p
l
e
t
o
s
a
m
p
l
e
.
*
F
i
x
e
d
a
t
v
a
l
u
e
o
f
1
.
0
t
o
d
e

n
e
t
h
e
m
e
t
r
i
c
o
f
t
h
e
s
e
c
o
n
d
-
o
r
d
e
r
l
a
t
e
n
t
v
a
r
i
a
b
l
e
i
n
t
h
e
u
n
s
t
a
n
d
a
r
d
i
z
e
d
s
o
l
u
t
i
o
n
.
ARCHITECTURE OF AGGRESSION 155
T
A
B
L
E
5
R
e
l
i
a
b
i
l
i
t
i
e
s
a
n
d
C
F
A
F
a
c
t
o
r
I
n
t
e
r
c
o
r
r
e
l
a
t
i
o
n
s
f
o
r
t
h
e
O
r
i
g
i
n
a
l
a
n
d
R
e

n
e
d
F
o
u
r
-
F
a
c
t
o
r
M
o
d
e
l
s
S
a
m
p
l
e
1
S
a
m
p
l
e
2
S
a
m
p
l
e
3
A
Q
f
a
c
t
o
r
s
M
o
d
e
l
P
A
V
A
A
N
G
H
O
P
A
V
A
A
N
G
H
O
P
A
V
A
A
N
G
H
O
P
A
O
r
i
g
i
n
a
l
8
6
a
8
4
a
8
4
b
R
e

n
e
d
8
0
a
8
0
a

V
A
O
r
i
g
i
n
a
l
6
0
7
4
a
6
4
6
8
a
6
2
7
5
b
R
e

n
e
d
5
7
7
7
a
6
3
7
3
a
5
8

A
N
G
O
r
i
g
i
n
a
l
7
4
8
1
8
2
a
7
2
8
2
8
3
a
3
7
3
7
8
0
b
R
e

n
e
d
6
3
7
6
7
1
a
6
6
7
4
7
3
a
3
9
3
5

H
O
O
r
i
g
i
n
a
l
5
4
4
7
6
3
7
6
a
5
2
4
8
5
7
8
1
a
7
5
6
4
6
7
8
3
b
R
e

n
e
d
5
3
5
3
5
7
7
3
a
4
7
4
3
4
3
7
7
a
8
0
6
6
8
5

N
o
t
e
:
P
A
ϭ
P
h
y
s
i
c
a
l
A
g
g
r
e
s
s
i
o
n
;
V
A
ϭ
V
e
r
b
a
l
A
g
g
r
e
s
s
i
o
n
;
A
N
G
ϭ
A
n
g
e
r
;
H
O
ϭ
H
o
s
t
i
l
i
t
y
.
S
a
m
p
l
e
1
ϭ
3
0
7
A
m
e
r
i
c
a
n
u
n
d
e
r
g
r
a
d
u
a
t
e
s
;
S
a
m
p
l
e
2
ϭ
2
0
0
B
r
i
t
i
s
h
u
n
d
e
r
g
r
a
d
u
a
t
e
s
(
A
r
c
h
e
r
,
H
o
l
l
o
w
a
y
,
&
M
c
L
o
u
g
h
l
i
n
,
1
9
9
5
)
;
S
a
m
p
l
e
3
ϭ
3
0
6
C
a
n
a
d
i
a
n
u
n
d
e
r
g
r
a
d
u
a
t
e
s
(
H
a
r
r
i
s
,
1
9
9
5
)
.
T
a
b
l
e
d
b
e
l
o
w
t
h
e
r
e
l
i
a
b
i
l
i
t
y
c
o
e
f

c
i
e
n
t
s
a
r
e
f
a
c
t
o
r
c
o
r
r
e
l
a
t
i
o
n
s
f
r
o
m
c
o
n

r
m
a
t
o
r
y
f
a
c
t
o
r
a
n
a
l
y
s
e
s
i
n
w
h
i
c
h
l
a
t
e
n
t
f
a
c
t
o
r
s
a
n
d
m
e
a
s
u
r
e
d
v
a
r
i
a
b
l
e
s
h
a
v
e
b
e
e
n
s
t
a
n
d
a
r
d
i
z
e
d
s
e
p
a
r
a
t
e
l
y
w
i
t
h
i
n
e
a
c
h
s
a
m
p
l
e
.
D
e
c
i
m
a
l
p
o
i
n
t
s
a
r
e
o
m
i
t
t
e
d
.
a
C
r
o
n
b
a
c
h

s
a
l
p
h
a
(
a
n
i
n
d
e
x
o
f
i
n
t
e
r
n
a
l
c
o
n
s
i
s
t
e
n
c
y
)
f
o
r
u
n
i
t
-
w
e
i
g
h
t
e
d
o
r
i
g
i
n
a
l
a
n
d
r
e

n
e
d
f
a
c
t
o
r
s
c
o
r
e
s
.
b
R
e
l
i
a
b
i
l
i
t
y
c
o
e
f

c
i
e
n
t
s
o
r
i
g
i
n
a
l
l
y
r
e
p
o
r
t
e
d
b
y
H
a
r
r
i
s
(
1
9
9
5
,
T
a
b
l
e
2
,
p
.
9
9
3
)
.
W
e
c
o
u
l
d
n
o
t
c
o
m
p
u
t
e
r
e
l
i
a
b
i
l
i
t
y
c
o
e
f

c
i
e
n
t
s
f
o
r
t
h
e
r
e

n
e
d
f
a
c
t
o
r
s
i
n
S
a
m
p
l
e
3
b
e
c
a
u
s
e
o
n
l
y
t
h
e
c
o
v
a
r
i
a
n
c
e
m
a
t
r
i
x
(
b
u
t
n
o
t
t
h
e
n
e
c
e
s
s
a
r
y
r
a
w
d
a
t
a
)
w
a
s
a
v
a
i
l
a
b
l
e
f
o
r
r
e
a
n
a
l
y
s
i
s
o
f
t
h
i
s
d
a
t
a
s
e
t
.
156 BRYANT AND SMITH
∆χ
2
(1)s Ͼ 3.8, ps Ͻ .05, and with Anger, both ∆χ
2
(1)s Ͼ 7.5, ps Ͻ .006,
compared to the American and British samples. Although interesting from
a cross-cultural perspective, these differences in factor interrelationships do
not alter the conclusion that the 12 AQ indicators measure the latent factors
in comparable ways for all three samples (cf. Kline, 1998).
Comparing the original and refined factors. For purposes of comparison,
Table 5 also displays the correlations among Buss and Perry’s (1992) original
four factors in these same three samples. As evident in this table, the pattern
of correlations among the refined factors is strikingly similar to the pattern
of correlations among the original Buss–Perry factors. Indeed, the refined
factors have only slightly lower internal consistency reliabilities than their
original counterparts. This is important because it suggests that refining the
factors improved the model’s overall goodness-of-fit, but did not substan-
tially reduce the reliabilities of the individual factors.
1
Testing a ‘‘Partially Disaggregated’’ form of Buss and Perry’s original
model. Before abandoning Buss and Perry’s (1992) original four-factor
model, we reconfigured it into a ‘‘partially disaggregated’’ measurement
model that preserved the full range of ‘‘totally disaggregated’’ item content
while reducing the number of indicators per factor (cf. Bagozzi & Edwards,
1998; Bagozzi & Heatherton, 1994; Hull et al., 1991). To do this, we modi-
fied the nine single-item indicators for PA into three composite measures,
using as indicators (a) the mean of the three PA items from the refined mea-
surement model; (b) the mean of original AQ items 1, 3, and 4; and (c) the
mean of original AQ items 5, 7, and 9. We modified the five single-item
indicators for VA into three measures, using as indicators (a) the mean of
the three VA items from the refined measurement model, (b) original AQ
item 10, and (c) original AQ item 12. We modified the seven single-item
indicators for ANG into three composite measures, using as indicators (a)
the mean of the three ANG items from the refined measurement model, (b)
the mean of original AQ items 16 and 17, and (c) the mean of original AQ
items 18 and 19. Finally, we modified the eight single-item indicators for
HO into three composite measures, using as indicators (a) the mean of the
three HO items fromthe refined measurement model; (b) the mean of original
AQ items 22, 26, and 27; and (c) the mean of original AQ items 28 and 29.
We then imposed this ‘‘partially disaggregated’’ four-factor model on the
1
As further evidence concerning the degree of conceptual overlap between the refined and
original Buss–Perry factors, we correlated unit-weighted factor scores for the former and the
latter within Samples 1 and 2. (We were unable to correlate factor scores in Sample 3 because
only the covariance matrix, and not the necessary raw data, was available for reanalysis.)
For Samples 1 and 2, respectively, these intercorrelations were uniformly high: (a) Physical
Aggression (.91 and .86), (b) Verbal Aggression (.90 in both samples), (c) Anger (.90 and
.85), and (d) Hostility (.83 and .85). These findings suggest that the refined AQfactors basically
measure the same latent constructs as the original factors.
ARCHITECTURE OF AGGRESSION 157
data of Samples 1 and 2 to evaluate its goodness-of-fit and cross-sample
generalizability. (We could not construct partially disaggregated indicators
for Sample 3 because only the covariance matrix was available for reanalysis
of this sample.) Although the partial disaggregation model provided a satis-
factory fit to the data of Sample 1, χ
2
(48, n ϭ 307) ϭ 157.6, χ
2
/df ϭ 3.3,
GFI ϭ .93, RMSEA ϭ .083, CFI ϭ .93, NNFI ϭ .91; it did not adequately
fit the data of Sample 2, χ
2
(48, n ϭ 200) ϭ 197.4, χ
2
/df ϭ 4.1, GFI ϭ
.87, RMSEA ϭ .120, CFI ϭ .88, NNFI ϭ .83. Multigroup analyses further
demonstrated that neither the factor loadings, ∆χ
2
(8, n ϭ 507) ϭ 32.2, p Ͻ
.0001, nor the factor variances–covariances, ∆χ
2
(10, n ϭ 507) ϭ 115.8,
p Ͻ .00001, of this partially disaggregated model were invariant across the
two samples. Thus, even in a ‘‘partially disaggregated’’ form, Buss and Per-
ry’s (1992) original four-factor model does not provide an acceptable mea-
surement model for the AQ.
Testing the Gender Invariance of the Refined Model. Does aggression
mean the same thing to men and women? Is the refined measurement model
for the AQ equally applicable to the data of males and females? This is a
critical question for researchers interested in using the AQ to compare levels
of aggression in men and women. With respect to this question, Buss and
Perry (1992) reported that men’s and women’s loadings differed in separate
four-factor PCA solutions, though they did not directly test the gender invari-
ance of the four-factor model.
The refined four-factor model generated gender-equivalent loadings for
the British Sample 2, ∆χ
2
(8, n ϭ200) ϭ6.0, p Ͼ.64, but not for the Ameri-
can Sample 1, ∆χ
2
(8, n ϭ 304) ϭ 21.4, p Ͻ .007. (Because only the pooled
covariance matrix was available for reanalysis of Harris’s data, we could not
test for gender invariance in Sample 3.) Follow-up multigroup CFAs re-
vealed that only one factor loading actually differed significantly for males
and females in Sample 1—the loading for Physical Aggression item 3
(‘‘There are people who have pushed me so hard that we came to blows’’)
was stronger for men than women, ∆χ
2
(8, n ϭ304) ϭ12.3, p Ͻ.0005—and
all other loadings for the four-factor model were gender-equivalent, ∆χ
2
(7,
n ϭ 304) ϭ 2.9, p Ͼ .89. These findings suggest that the loadings of the
refined four-factor model are largely invariant with respect to gender. Like-
wise for the hierarchical model, the overarching second-order Aggression
factor showed comparable relationships with the four first-order factors for
males and females in both Sample 1, ∆χ
2
(4, n ϭ 304) ϭ 1.5, p Ͼ .82, and
Sample 2, ∆χ
2
(4, n ϭ 200) ϭ 5.9, p Ͼ .20. Thus, the refined factors appear
to have substantially the same meaning for men and women.
Stage Three: Assessing Construct Validity
What evidence of convergent or discriminant validity is there for the re-
fined AQ factors? And how does this validity evidence compare to that for
158 BRYANT AND SMITH
Buss and Perry’s original AQ factors? To address these questions, we used
CFA with equality constraints in Sample 1 data to test hypotheses about
relationships between (a) the four AQ factors in both their refined and origi-
nal forms; and (b) the four criterion measures of physical assault, verbal
hostility, anger arousal, and global hostility. We used CFA rather than tradi-
tional correlational methods because it allowed us to estimate the relationship
between aggression subtraits and criterion measures, partialing out measure-
ment error, and also provided a way to systematically test hypotheses about
the strength of associations across measures.
Because measurement error attenuates relationships, different measures
may demonstrate different interrelationships due to differences in reliability.
For this reason, it is important to control for differential reliability when
assessing the strength of relationships between observed measures. Tradi-
tional correlational methods assume that all analyzed variables are measured
perfectly and therefore do not allow researchers to adjust for differential
reliability (Kline, 1998; Maruyama, 1998). Using CFA, in contrast, enabled
us to control for differences in the reliabilities of the AQ factors and the
criterion measures, which might otherwise influence the strength of the ob-
served associations (Bagozzi, 1993; Judd, Jessor, & Donovan, 1986).
Another advantage of CFA over the traditional correlational approach is
that it allowed us to use equality constraints to systematically test hypotheses
about the strength of the relationships among the latent constructs. Typically,
researchers have simply eyeballed differences in correlation coefficients to
determine the degree to which measures show convergent or discriminant
validity. Using CFA, in contrast, enabled us to test the statistical significance
of differences in the magnitude of validity coefficients by contrasting the
goodness-of-fit chi-square values (and degrees of freedom) of two nested
models: one that constrained the correlations in question to have equal value
and one that contained no equality constraint. A significant difference in
these two nested chi-square values signifies that the correlations differ in
magnitude (cf. Bryant & Baxter, 1997; Bryant & Yarnold, 1995).
To obtain the multiple indicators required for CFA while also minimizing
the number of measured variables in the model, we used the ‘‘partial disag-
gregation’’ approach to parcel each of the four criterion measures into two
composite indicators. For the Buss–Durkee Physical Assault scale, we
summed responses to the odd-numbered items (1, 3, and 5) to create one
indicator and summed responses to the even-numbered items (2 and 4) to
create a second indicator. For the Buss–Durkee Verbal Hostility scale, we
summed responses to items 1–4 to create one indicator and summed re-
sponses to items 5–8 to create a second indicator. For the MAI Anger
Arousal scale, we summed responses to items 1 and 2 to create one indicator
and summed responses to items 3 and 4 to create a second indicator. Total
scores on split-halves of the Cook–Medley Hostility Scale served as indica-
ARCHITECTURE OF AGGRESSION 159
TABLE 6
Correlations of the Refined and Original AQ Factors with the Criterion Measures
Refined AQ factors Original AQ factors
Criterion measures PA VA ANG HO PA VA ANG HO
Buss–Durkee Physical Assault 85 62 65 43 90 62 70 44
Buss–Durkee Verbal Hostility 57 64 72 45 63 66 74 45
MAI Anger Arousal 64 66 91 77 64 64 93 88
Cook–Medley Hostility 46 60 64 89 49 58 65 84
Note. PA ϭPhysical Aggression; VA ϭVerbal Aggression; ANG ϭAnger; HO ϭ Hostil-
ity. These data are from a random subset of 180 American undergraduates (70 males and 110
females) from Sample 1. Tabled are standardized φ coefficients (analogous to Pearson correla-
tion coefficients) from confirmatory factor analyses. These coefficients reflect the degree of
association between latent constructs that have been adjusted for differences in measurement
reliability. Cronbach’s αs for unit-weighted factor scores on the criterion measures were as
follows: Physical Assault (.64), Verbal Hostility (.63), Anger Arousal (.82), and Cook-Medley
Hostility (.86). All φs are statistically significant at p Ͻ .00001, two-tailed.
tors for the criterion measure of Hostility. This yielded eight composite in-
dicators of four criterion factors. CFA revealed that the intended four-
factor model for the criterion measures (i.e., correlated factors of Physical
Assault, Verbal Hostility, Anger Arousal, and Global Hostility) was an ac-
ceptable measurement model for the eight composite indicators, χ
2
(14, n ϭ
180) ϭ 19.2, p Ͼ .16, GFI ϭ .97, RMSEA ϭ .0454, CFI ϭ .99, NNFI ϭ
.98.
Refined AQ Factors. We next analyzed the 12 items constituting the re-
fined four-factor model for the AQ together with the eight criterion measures
and examined the factor intercorrelations. Table 6 presents the correlations
between the four refined AQ factors and the four criterion constructs for
this CFA model. To assess discriminant validity, we first examined these
correlations separately within each factor, conducting an initial omnibus test
of the homogeneity of correlations across criterion measures for each column
in the table. The hypothesis of equality in correlations across criteria (i.e.,
no discriminant validity) was rejected for the AQ factors of Physical Aggres-
sion, ∆χ
2
(3, n ϭ 180) ϭ 22.5, p Ͻ .0001; Anger, ∆χ
2
(3, n ϭ 180) ϭ 23.0,
p Ͻ .0001; and Hostility, ∆χ
2
(3, n ϭ 180) ϭ 50.0, p Ͻ .0001; but not for
Verbal Aggression, ∆χ
2
(3, n ϭ 180) ϭ 1.1, p Ͼ .77.
Supporting the construct validity of the Physical Aggression factor, fol-
low-up CFA tests using equality constraints revealed that levels of this AQ
factor were more strongly correlated with the Physical Assault criterion than
with the criteria of Verbal Hostility, ∆χ
2
(1, n ϭ 180) ϭ 10.0, p Ͻ .002;
Anger Arousal, ∆χ
2
(1, n ϭ180) ϭ6.6, p Ͻ.01; and Global Hostility, ∆χ
2
(1,
n ϭ 180) ϭ 19.2, p Ͻ .0001. Supporting the construct validity of the Anger
160 BRYANT AND SMITH
factor, levels of this AQ factor were more strongly correlated with the Anger
Arousal criterion than with the criteria of Verbal Hostility, ∆χ
2
(1, n ϭ 180)
ϭ 9.2, p Ͻ .025; Anger Arousal, ∆χ
2
(1, n ϭ 180) ϭ 6.8, p Ͻ .009; and
Global Hostility, ∆χ
2
(1, n ϭ 180) ϭ 20.7, p Ͻ .0001. And supporting the
construct validity of the Hostility factor, levels of this AQ factor were more
strongly correlated with the Global Hostility criterion than with the criteria
of Physical Aggression, ∆χ
2
(1, n ϭ 180) ϭ 13.3, p Ͻ .003; Verbal Aggres-
sion, ∆χ
2
(1, n ϭ 180) ϭ 14.0, p Ͻ .0002; and Anger Arousal, ∆χ
2
(1, n ϭ
180) ϭ 6.0, p Ͻ .015. Thus, three of the four refined AQ factors showed
evidence of convergent and discriminant validity.
2
Original AQ Factors. We next analyzed the 29 items constituting Buss
and Perry’s (1992) original four-factor measurement model together with
the eight criterion measures and examined the factor intercorrelations. Table
6 also presents the correlations between the four original AQ factors and the
four criterion constructs for this CFA model. As with the refined measure-
ment model, we first examined these correlations separately within factors,
conducting an initial omnibus test of the homogeneity of correlations for
each column in the table. Paralleling results for the refined AQ factors, the
hypothesis of equality in correlations across criteria (i.e., no discriminant
validity) was rejected for the original AQ factors of Physical Aggression,
∆χ
2
(3, n ϭ 180) ϭ 27.4, p Ͻ .0001; Anger, ∆χ
2
(3, n ϭ 180) ϭ 31.1, p Ͻ
.0001; and Hostility, ∆χ
2
(3, n ϭ 180) ϭ 50.6, p Ͻ .0001; but not for Verbal
Aggression, ∆χ
2
(3, n ϭ 180) ϭ 1.5, p Ͼ .68. Thus, in both refined and
original forms, Verbal Aggression lacked discriminant validity.
Supporting the construct validity of the original Physical Aggression fac-
tor, follow-up CFA tests using equality constraints revealed that levels of
2
We also assessed the discriminant validity of unit-weighted factor scores for the refined
AQ model in terms of their predictive utility in distinguishing males and females. Buss and
Perry (1992, p. 455) originally reported that males had higher scores than females on Physical
Aggression (PA), Verbal Aggression (VA), and Hostility (HO), but not on Anger (ANG).
Multivariate analyses of variance revealed a significant multivariate main effect of gender for
Sample 1, F(4, 299) ϭ 21.1, p Ͻ .0001; Sample 2, F(4, 195) ϭ 4.3, p Ͻ .002; Sample 4,
F(4, 166) ϭ 6.5, p Ͻ .0001; and Sample 5, F(4, 165) ϭ 8.9, p Ͻ .0001. (We could not test
for gender differences in factor means for Sample 3 because only the covariance matrix was
available for reanalysis of these data.) Univariate ANOVAs revealed that males scored higher
than females on (a) PA in American Samples 1, 4, and 5, all ps Ͻ .0001, effect sizes d ϭ
.85, .54, and 1.0, respectively, but not in British Sample 2, p Ͼ .60, d ϭ .07; (b) VA in
Samples 1 and 4, ps Ͻ .02, ds ϭ .45 and .42, respectively, but not in Samples 2 and 5,
ps Ͼ .17, ds ϭ .11 and .23, respectively; (c) ANG in Sample 5, p Ͻ .02, d ϭ .58, but not
in Samples 1, 2, and 4, ps Ͼ .37, ds ϭ .15, .12, and .03, respectively; and (d) HO in Samples
1 and 5, ps Ͻ .02, ds ϭ .35 and .46, respectively, but not in Samples 2 and 4, ps Ͼ .07, ds
ϭ.25 and .22, respectively. Though somewhat inconsistent across samples, these results gener-
ally converge with those of previous researchers and provide at least partial support for the
discriminant validity of the refined AQ factors.
ARCHITECTURE OF AGGRESSION 161
this AQ factor were more strongly correlated with the Physical Assault crite-
rion than with the criteria of Verbal Hostility, ∆χ
2
(1, n ϭ 180) ϭ 11.7,
p Ͻ .0007; Anger Arousal, ∆χ
2
(1, n ϭ 180) ϭ 11.9, p Ͻ .0006; and Global
Hostility, ∆χ
2
(1, n ϭ 180) ϭ 26.1, p Ͻ .0001. Supporting the construct
validity of the Anger factor, levels of this AQ factor were more strongly
correlated with the Anger Arousal criterion than with the criteria of Verbal
Hostility, ∆χ
2
(1, n ϭ180) ϭ8.7, p Ͻ.004; Anger Arousal, ∆χ
2
(1, n ϭ180)
ϭ 8.3, p Ͻ .004; and Global Hostility, ∆χ
2
(1, n ϭ 180) ϭ 30.2, p Ͻ .0001.
And partially supporting the construct validity of the Hostility factor, levels
of this AQ factor were more strongly correlated with the Global Hostility
criterion than with the criteria of Physical Assault, ∆χ
2
(1, n ϭ 180) ϭ 20.0,
p Ͻ .0001, and Verbal Hostility, ∆χ
2
(1, n ϭ 180) ϭ 24.7, p Ͻ .0001. How-
ever, unlike the refined Hostility factor, Buss and Perry’s original Hostility
factor correlated equally with both the Global Hostility and Anger Arousal
criteria, ∆χ
2
(1, n ϭ180) ϭ0.7, p Ͼ.41. This result is consistent with earlier
findings of a high degree of overlap between these two dimensions (Harris,
1997). Thus, only two of the four original AQ factors showed evidence of
convergent and discriminant validity.
Considered together, what do these results tell us about the refined mea-
surement model? First, the construct validity of the refined AQ factors ap-
pears to be as good or better than the construct validity of the original Buss–
Perry factors. For both models, the Physical Aggression and Anger factors
showed the strongest evidence of convergent and discriminant validity,
whereas the Verbal Aggression factor showed the weakest. The Hostility
factor, in contrast, had stronger discriminant validity in its refined form than
in its original form. Taken as a whole, these findings support the construct
validity of the refined measurement model for the AQ, and they suggest that
the process of modifying the original factors has essentially preserved their
conceptual content.
Stage Four: Comparing the Original and Short Forms of the AQ
Our final research objective was to readminister the refined subset of 12
AQ items as a shortened version of the AQ and to evaluate the degree to
which the four-factor measurement model replicated in this new form (cf.
Yarnold, Bryant, & Grimm, 1987). Accordingly, we analyzed the responses
of Samples 4 and 5 to this new, short form of the AQ. Results revealed that
the refined four-factor model replicated in both samples. Specifically, the
intended four-factor model provided an acceptable measurement model for
the 12 items in both Sample 4, χ
2
(48, n ϭ 171) ϭ 116.94, GFI ϭ .90,
RMSEA ϭ .092, CFI ϭ.92, NNFI ϭ .90; and Sample 5, χ
2
(1, n ϭ 170) ϭ
119.14, GFI ϭ .90, RMSEA ϭ .094, CFI ϭ .91, NNFI ϭ .88. The refined
hierarchical also provided an acceptable measurement model for the 12 items
in both Sample 4, χ
2
(50, n ϭ 171) ϭ 118.1, GFI ϭ .89, RMSEA ϭ .090,
162 BRYANT AND SMITH
CFIϭ.92, IFI ϭ .93, NNFI ϭ .90; and Sample 5, χ
2
(50, n ϭ 170) ϭ 119.4,
GFI ϭ .90, RMSEA ϭ .091, CFI ϭ .92, NNFI ϭ .89.
As a measure of each factor’s internal consistency, we also computed
Cronbach’s alphas for unit-weighted scores on the refined AQ factors. For
both Samples 4 and 5, respectively, reliability coefficients were generally
acceptable for the refined factors of Physical Aggression (.79 and .80), Ver-
bal Aggression (.83 and .80), Anger (.76 and .76), and Hostility (.75 and
.70). These reliabilities are generally comparable to those for Samples 1 and
2, which used the longer, original form of the AQ (see Table 5).
We also used multigroup CFA to assess the generalizability of the refined
measurement model across Samples 4 and 5 and across all five samples.
Confirming replicability, the refined four-factor model produced equivalent
loadings, ∆χ
2
(8, n ϭ 341) ϭ 4.7, p Ͼ .78, and equivalent factor variances–
covariances, ∆χ
2
(10, n ϭ341) ϭ5.8, p Ͼ.82, for Samples 4 and 5. Analyz-
ing the data of all five samples simultaneously, we found that the refined
four-factor model produced (a) fully invariant loadings in Samples 1, 2,
4, and 5; and (b) 11 of 12 invariant loadings in Sample 3, ∆χ
2
(46, n ϭ1154)
ϭ24.9, p Ͼ.99. In addition, the overarching higher-order Aggression factor
had the same relationships with the first-order AQ factors across Samples
1, 2, 4, and 5, ∆χ
2
(12, n ϭ848) ϭ14.5, p Ͼ .26. Thus, the refined measure-
ment model replicates with the shortened version of the AQ.
As a strong test of the comparability of the factors represented in the short
and long forms of the AQ, we reran the analyses for Samples 4 and 5 fixing
the factor intercorrelations in the model to the exact values found earlier for
our American Sample 1 (see Table 5). Using the factor correlations from
the earlier sample did not significantly worsen the fit of the model for either
Sample 4, ∆χ
2
(6, n ϭ 171) ϭ 12.4, p Ͼ .05, or Sample 5, ∆χ
2
(6, n ϭ
170) ϭ 11.8, p Ͼ .06. Thus, shortening the AQ to 12 items does not appear
to change the conceptual meaning of the underlying aggression subtraits.
GENERAL DISCUSSION
The most important finding of the present study is the discovery of an
appropriate measurement model for the Aggression Questionnaire. Although
Buss and Perry’s (1992) original four-factor model has a strong theoretical
foundation, it explains too little common variance in the 29 AQ items to serve
as a formal measurement model. The refined model, in contrast, preserves the
conceptual content of the original model, but improves its statistical preci-
sion. That the refined model replicates with both the original and short ver-
sions of the AQ, is equally applicable for both males and females, and holds
across independent samples from three different countries (i.e., United
States, England, and Canada) increases our confidence in the external valid-
ity of this measurement model.
Our data also provide strong support for the convergent and discriminant
ARCHITECTURE OF AGGRESSION 163
validity of the refined Physical Aggression, Anger, and Hostility factors.
Each of these factors correlated more strongly with the criterion measure to
which it is presumed to correspond than to the other criterion measures. The
Verbal Aggression factor, in contrast, showed no discriminant validity in
either its original or refined forms, and was at least moderately correlated
with each of the other AQ factors. Clearly, future work is needed to establish
the construct validity of this particular subtrait.
Our data further demonstrate that the refined Hostility factor has greater
discriminant validity than the original Hostility factor. The latter correlated
equally with scores on both the Cook–Medley Hostility Scale and the MAI
Anger Arousal subscale, whereas the former correlated more strongly with
Cook–Medley scores than with the other criterion measures. It is instructive
to note that Buss and Perry (1992) originally defined hostility as the cognitive
component of aggression, consisting of ‘‘feelings of ill will and injustice’’
(p. 457). By this definition, hostility is a cognitive dimension that includes
negative feelings. That the original Hostility factor had equivalent correla-
tions with the hostility and anger criteria is entirely consistent with this con-
ceptualization. The refined Hostility factor, in contrast, excludes items tap-
ping jealousy (item 22), paranoia (item 28), and suspiciousness (items 27
and 29). Omitting these items appears to disentangle Hostility from Anger,
giving the former a more exclusively cognitive focus.
What measurement model should researchers adopt in using the AQ, and
how should they best score this instrument? Our findings clearly indicate
that dispositional aggression (as measured by the AQ) is multidimensional.
Thus, examining the four separate subtraits is more reasonable both concep-
tually and statistically than relying solely on a pooled ‘‘total score.’’ Indeed,
by collapsing across multiple dimensions, AQ total score could conceivably
obscure differences that exist for individual factors. This suggests that it
would be unwise for researchers to use only the 29-item total score in quanti-
fying responses to the AQ.
Our data also demonstrate that the refined four-factor model is superior
to Buss and Perry’s (1992) original four-factor model in its overall goodness-
of-fit to the data. In other words, the 12 AQ items constituting the refined
model more closely reflect the underlying subtraits of physical aggression,
verbal aggression, anger, and hostility than do the 29 items constituting the
original model. Although the four original factors offer somewhat greater
internal consistency in terms of Cronbach’s alpha, this apparent advantage
disappears when the reliabilities of the refined factors are corrected for differ-
ences in the number of constituent items (adjusted αs ϭ.88 Ϫ.92), using the
Spearman–Brown prophecy formula (Magnusson, 1967). Thus, the refined
model appears to provide a valid and reliable means of measuring the four
aggression subtraits, using either the original 29-item AQ or its 12-item short
form.
164 BRYANT AND SMITH
Whether one should adopt the four-factor model or its hierarchical coun-
terpart depends on one’s research objectives. The first-order model enables
one to examine the relationships among each of the four AQ factors as well
as the specific correlates of each subtrait of aggression. The higher-order
model, in contrast, enables one to examine the common variance among the
four AQ factors as an independent or dependent variable in its own right.
In either case, the present study suggests that researchers can enhance both
conceptual and predictive precision by using the refined measurement model
for the AQ.
Our research conclusions are constrained by several important limitations.
First, we used only self-report measures as criteria in assessing construct
validity. Clearly, it would be advantageous to include behavioral and physio-
logical criterion measures of physical and verbal aggression, anger, and hos-
tility (cf. Leibowitz, 1968). This would increase confidence that the AQ actu-
ally measures these aggression subtraits. In addition, we used only college
student samples across the three countries, and our results may not generalize
to older adults or offender populations (cf. Williams et al., 1996).
In closing, we note two unresolved measurement issues concerning the
AQ. The first issue concerns the nature of the response scale used to quantify
AQ responses. Although we changed the original 5-point scale to a 6-point
scale in the AQ short form, we preserved the original anchors for the scale
endpoints: extremely uncharacteristic of me to extremely characteristic of
me. However, the exact meaning of this scale remains somewhat unclear.
Specifically, is an uncharacteristic–characteristic scale unipolar or bipolar
(cf. Sudman & Bradburn, 1982)? Might respondents interpret ‘‘extremely
uncharacteristic’’ to mean that the particular characteristic in question is so
unlike them that its opposite actually holds true for them? Such bipolarity
would be conceptually inconsistent with traditional theoretical formulations
of aggression. In the future, it might be better to anchor the scale with not
at all characteristic of me to very much characteristic of me, so as to avoid
conceptual confusion.
A final measurement issue concerns the order of the items constituting
the AQ. Buss and Perry (1992) did not report the order of the 29 items in
the original AQ, but simply advised researchers to ‘‘make up their own . . .
version by scrambling the items so that items do not pile up’’ (footnote 1,
p. 453). Accordingly, different users have used different randomized order-
ings of the AQ items. This approach has drawbacks, however. Though appar-
ently intended to enhance measurement validity, the practice of rerandomiz-
ing the order of the AQ items across studies makes it extremely difficult to
conduct secondary analyses of existing AQ data sets, as done in the present
study. When comparing data across studies, one must not only determine
the specific order of the items that each researcher has used, but must also
match up items across studies and then rearrange all AQ items in each data
ARCHITECTURE OF AGGRESSION 165
set into one universal order, constructing a complex formof ‘‘Rosetta Stone’’
in the process. For example, AQ item 11 in our Sample 1 is (a) item 21 in
Buss and Perry’s (1992, p. 454) Table 1; (b) item 22 in Archer, Holloway,
and McLoughlin’s (1995) data set; (c) item 29 in Harris’s (1995) data set;
and (d) item 5 in our 12-item short form of the AQ. Rerandomizing the order
of AQitems at each administration makes little sense, given the methodologi-
cal sufficiency of a single random order. To facilitate comparison of results
across studies, we strongly urge future researchers to adopt a standard ran-
dom order of the AQ items (the note for Table 2 reports the order we have
used for both the original and short forms of the AQ).
Finally, we alert researchers to the fact that Western Psychological Ser-
vices (Los Angeles, CA) has recently developed a new 34-item version of
the original 29-item AQ (Buss & Warren, 2000). This new instrument con-
tains all but one of the original Buss–Perry items (14 of which have been
modified to improve readability), along with a new, 6-item Indirect Aggres-
sion subscale. Although Indirect Aggression was initially a component of
the Buss–Durkee Hostility Inventory (Buss & Durkee, 1957), Buss and Perry
(1992) omitted this dimension of aggression in their original AQ. Research-
ers wishing to use the short form of the AQ should note that the questions
in the new AQ have been arranged so that the first 12 items represent the
indicators constituting our refined measurement model. Advantages to using
this new 34-item AQ (Buss & Warren, 2000), as opposed to the original 29-
item AQ (Buss & Perry, 1992), include the availability of (a) more extensive
evidence supporting construct validity; (b) large-scale, national norms; (c)
a reliable subscale to tap Indirect Aggression; and (d) a single, fixed order
for the AQ items.
REFERENCES
Archer, J., Kilpatrick, G., & Bramwell, R. (1995). Comparison of two aggression inventories.
Aggressive Behavior, 21, 371–380.
Archer, J., Holloway, R., & McLoughlin, K. (1995). Self-reported physical aggression among
young men. Aggressive Behavior, 21, 325–342.
Bagozzi, R. P. (1993). Assessing construct validity in personality research: Applications to
measures of self-esteem. Journal of Research in Personality, 27, 49–87.
Bagozzi, R. P., & Edwards, J. R. (1998). A general approach for representing constructs in
organizational research. Organizational Research Methods, 1, 45–87.
Bagozzi, R. P., & Heatherton, T. F. (1994). A general approach to representing multifaceted
personality constructs: Application to state self-esteem. Structural Equation Modeling,
1, 35–67.
Barefoot, J. C., Dahlstrom, W. G., & Williams, R. B., Jr. (1983). Hostility, CHD incidence
and total mortality: A 25-year follow-up study of 255 physicians. Psychosomatic Medi-
cine, 45, 59–63.
Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological Bulletin,
107, 238–246.
166 BRYANT AND SMITH
Bentler P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis
of covariance structures. Psychological Bulletin, 88, 588–606.
Bernstein, I. A., & Gesn, P. R. (1997). On the dimensionality of the Buss/Perry Aggression
Questionnaire. Behavior Research and Therapy, 35, 563–568.
Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley.
Browne, M. W., & Cudeck, R. (1989). Single-sample cross-validation indices for covariance
structures. Multivariate Behavioral Research, 24, 445–455.
Bryant, F. B., & Baxter, W. J. (1997). The structure of positive and negative automatic cogni-
tion. Cognition and Emotion, 11, 225–258.
Bryant, F. B., & Yarnold, P. R. (1995). Principal-components analysis and exploratory and
confirmatory factor analysis. In L. G. Grimm & P. R. Yarnold (Eds.), Reading and under-
standing multivariate statistics (pp. 99–136). Washington, DC: American Psychological
Association.
Bryant, F. B., Yarnold, P. R., & Grimm, L. G. (1996). Toward a measurement model of the
Affect Intensity Measure: A three-factor structure. Journal of Research in Personality,
30, 223–247.
Byrne, B. M., Shavelson, R. J., & Muthen, B. (1989). Testing for the equivalence of factor
covariance and mean structures: The issue of partial measurement invariance. Psychologi-
cal Bulletin, 105, 456–466.
Buss, A. H. (1961). The psychology of aggression. New York: Wiley.
Buss, A. H., & Durkee, A. (1957). An inventory for assessing different kinds of hostility.
Journal of Consulting Psychology, 21, 343–349.
Buss, A. H., Fischer, H., & Simmons, A. J. (1962). Aggression and hostility in psychiatric
patients. Journal of Consulting, 26, 84–89.
Buss, A. H., & Perry, M. (1992). The Aggression Questionnaire. Journal of Personality and
Social Psychology, 63, 452–459.
Buss, A. H., & Warren, W. L. (2000). The Aggression Questionnaire manual. Los Angeles:
Western Psychological Services.
Cook, W. W., & Medley, D. M. (1954). Proposed Hostility and Pharisaic-Virtue scales for
the MMPI. Journal of Applied Psychology, 38, 414–418.
Felsten, G., & Hill, V. (1998). Aggression Questionnaire hostility scale predicts anger in re-
sponse to mistreatment. Behavior Research and Therapy, 37, 87–97.
Harburg, E., Erfurt, J. C., Hauenstein, L. S., Chape, C., Schull, W. J., & Schork, M. A. (1973).
Socio-ecological stress, suppressed hostility, skin color, and black–white male blood
pressure. Psychosomatic Medicine, 35, 276–296.
Harris, J. A. (1995). Confirmatory factor analysis of the Aggression Questionnaire. Behavior
Research and Therapy, 33, 991–993.
Harris, J. A. (1997). A further evaluation of the Aggression Questionnaire: Issues of validity
and reliability. Behavior Research and Therapy, 35, 1047–1053.
Hoelter, J. W. (1983). The analysis of covariance structures: Goodness-of-fit indices. Sociolog-
ical Methods and Research, 11, 325–344.
Hu, L.-T., & Bentler, P. M. (1998). Fit indices in covariance structure modeling: Sensitivity
to underparameterized model misspecification. Psychological Methods, 3, 424–453.
Hull, J. G., Lehn, D. A., & Tedlie, J. C. (1991). A general approach to testing multifaceted
personality constructs. Journal of Personality and Social Psychology, 61, 932–945.
Joreskog, K. G., & Sorbom, D. G. (1996). LISREL 8: User’s reference guide. Chicago: Scien-
tific Software.
ARCHITECTURE OF AGGRESSION 167
Judd, C. M., Jessor, R., & Donovan, J. E. (1986). Structural equation models and personality
research. Journal of Personality, 54, 149–198.
Kline, R. B. (1998). Principles and practice of structural equation modeling. New York:
Guilford.
Leibowitz, G. (1968). Comparison of self-report and behavioral techniques of assessing ag-
gression. Journal of Consulting and Clinical Psychology, 32, 21–25.
MacCallum, R. C. (1986). Specification searches in covariance structure modeling. Psycholog-
ical Bulletin, 100, 107–120.
MacCallum, R. C., Roznowski, M., & Necowitz, L. B. (1992). Model modification in covari-
ance structure analysis: The problem of capitalization on chance. Psychological Bulletin,
111, 490–504.
Magnusson, D. (1967). Test theory. Reading, MA: Addison–Wesley.
Maiuro, R. D., Cahn, T. S., Vitaliano, P. P., Wagner, B. C., & Zegree, J. B. (1988). Anger,
hostility, and depression in domestically violent versus generally assaultive men and non-
violent control subjects. Journal of Consulting and Clinical Psychology, 56, 17–23.
Marsh, H. W., Balla, J. R., & Hau, K.-T. (1996). An evaluation of incremental fit indices: A
clarification of mathematical and empirical properties. In G. A. Marcoulides & R. E.
Schumacker (Eds.), Advanced structural equation modeling: Issues and techniques
(pp. 315–353). Mahwah, NJ: Erlbaum.
Maruyama, G. M. (1998). Basics of structural equation modeling. Thousand Oaks, CA: Sage.
McDonald, R. P., & Marsh, H. W. (1990). Choosing a multivariate model: Noncentrality and
goodness of fit. Psychological Bulletin, 107, 247–255.
Meesters, C., Muris, P., Bosma, H., Schouten, E., & Beuving, S. (1996). Psychometric evalua-
tion of the Dutch version of the Aggression Questionnaire. Behavior Research and Ther-
apy, 34, 839–843.
Nichols, J. G., Licht, B. G., & Pearl, R. A. (1982). Some dangers of using personality question-
naires to study personality. Psychological Bulletin, 92, 572–580.
Novaco, R. (1975). Anger control: The development and evaluation of an experimental treat-
ment. Lexington, MA: Lexington Books.
Shekelle, R. B., Gale, M., Ostfeld, A. M., & Paul, O. (1983). Hostility, risk of coronary heart
disease, and mortality. Psychosomatic Medicine, 45, 109–114.
Siegel, J. M. (1986). The Multidimensional Anger Inventory. Journal of Personality and Social
Psychology, 51, 191–200.
Steiger, J. H. (1990). Structural model evaluation and modification: An interval estimation
approach. Multivariate Behavioral Research, 25, 173–180.
Sudman, S., & Bradburn, N. M. (1982). Asking questions: A practical guide to questionnaire
construction. San Francisco: Jossey–Bass.
Tucker, L. R., & Lewis, C. (1973). A reliability coefficient for maximum likelihood factor
analysis. Psychometrika, 38, 1–10.
Velicer, W. F., Govia, J. M., Cherico, N. P., & Corriveau, D. P. (1985). Item format and the
structure of the Buss–Durkee Hostility Inventory. Aggressive Behavior, 11, 65–82.
Williams, T. W., Boyd, J. C., Cascardi, M. A., & Poythress, N. (1996). Factor structure and
convergent validity of the Aggression Questionnaire in an offender population. Psycho-
logical Assessment, 8, 398–403.
Yarnold, P. R., Bryant, F. B., & Grimm, L. G. (1987). Comparing the long and short forms
of the Student Jenkins Activity Survey. Journal of Behavioral Medicine, 10, 75–90.
Zillmann, D. (1979). Hostility and aggression. Hillsdale, NJ: Earlbaum.